Trick or treat: Children's understanding of surprise

Cognitive Development, 8, 27-46 (1993)

Trick or Treat: Children's Understanding of Surprise

Rick MacLaren David Olson

The Ontario Institute for Studies in Education

This study examines the hypothesis that an understanding of false belief would lead to a radical change in young children's understanding of surprise. In Experi- ment 1, children aged 3 to 8 years were asked to assess the knowledge state of another person and to then choose an object that would surprise that person. The results showed that whereas the 3-year-olds' choice of surprising object varied with the object, the 5-year-olds' choice of object varied with their assessment of the other's knowledge state. Hence, understanding surprise depends on an understanding of false belief. In Experiment 2, the number of questions was reduced and children were required to match a schematized facial expression to the object judged to be surprising. Again, older children, unlike their younger counterparts, pointed out that surprised faces are made when another's expectations are violated. Once children begin to ascribe belief states to others they begin to understand that surprise depends upon the unexpected. The results help resolve the 'differences in the findings of Wellman and Banerjee (1991) and Hadwin and Perner (1991) on children's understanding of surprise. In natural judgements, young children employ a principle of desirability; older children employ principles of belief violation.

It is widely held that understanding false bel ief is a crucial developmenta l

miles tone in the chi ld ' s deve lop ing theory o f mind (for a summary see Ast ington,

Harris, & Olson, 1988). The 4-year-old child begins to understand that another 's

behavior depends upon the other 's beliefs about the world, whether true or false.

This is contrasted with the ability of the 3-year-old, who can predict another 's

behavior as long as it is congruent with the true state of affairs in the world, but

The authors would like to thank Earl Sweeney and Margot Mather. as well as the staff and children at the Ontario Science Centre and the OSCOTT program for their participation in this study. The authors would also like to thank Lynd Forguson. Elana Joram, and Hal White for comments on an earlier draft of the manuscript, and Sue Elgie and Bill Postl for statistical advice. The research for Experiment 1 was conducted as part of Rick MacLaren's Masters thesis, funded by NSERC support to David Olson.

Correspondence and requests for reprints should be sent to Richard fan MacLaren, Department of Psychology, Lakehead University, 955 Oliver Road. Thunder Bay. Ontario P7B 5El.

Manuscript received May 9, 1991; revision accepted July 21, 1992 27

28 Rick MacLaren and David Olson

cannot understand that another's behavior might be based on a false belief about the world• Accordingly, we would expect that if the false-belief task is the litmus test for a certain level of development of a theory of mind, then other mental- state concepts might be premised on an ability to understand false belief in the self and others• Surprise, we propose, is one such concept•

The false-belief task employed by Wimmer and Perner ( 1983; Hogrefe, Wim- mer, & Perner, 1986; Perner, 1988; Perner, Leekam, & Wimmer, 1987), was designed to test children's competence at representing another's belief when that belief was different from what the subject knew to be true. In that study, children were told a story about a character named Maxi, who puts chocolate into a BLUE cupboard. In Maxi's absence, the chocolate is moved to a GREEN cupboard. Children are then asked where Maxi will look for the chocolate. If subjects have an explicit representation of Maxi's belief, then they can use the belief as a means for interpreting and predicting Maxi's speech and thought--predicting that Maxi will look in the BLUE cupboard, think it is in the BLUE cupboard, or say it is in the BLUE cupboard. In understanding and explaining actions, 4- and 5-year- olds, but not 3-year-olds, appealed to the other person's beliefs, not to the external world.

Is this new understanding of belief related to changes in understanding of surprise? Wellman and Bartsch (1988) examined 4-year-old children's ratings of happiness and surprise in story contexts• They argued that happiness and surprise depend on prior mental states; happiness results when one gets what one wanted, and surprise results when one gets what was not expected• The results of their experiment confirmed this hypothesis• Specifically, 4-year-old children rated as surprising those stories in which an occurring event violated a held belief.

However, Wellman and Bartsch's experiment does not address two issues, one methodological and one theoretical. First, Wellman and Bartsch used a variety of story scripts (each with two different outcomes) as stimulus materials, but they did not discuss the results for individual story outcomes• Hence, it is not clear that subjects would deem outcomes as surprising if they were also undesirable or uninteresting. Young children, we speculate, might have based their judgements on the presence of interesting or novel objects rather than on the character's antecedent mental state. In fact, Wellman and Bartsch (1988) concluded that desirability was a factor in young children's judgements:

• . . in spite of the statement that the protagonist liked both outcomes in Think stories, children found it difficult to accept that a chm'acter with a definite belief about an object or event had no concomitant desire for it. They thus attributed an unmentioned desire to the character as well. (p. 253)

Hence, it is possible that these subjects believed that to be surprising an object had to be desirable in addition to being unexpected. In this study, this factor is manipulated such that some surprises involve unexpected treats whereas others involve failed expectations (i.e., disappointments)•

Children's Understanding of Surprise 29

Second, Wcllman and Bartsch did not address whether the development of an understanding of surprise precedes, comes with, or succeeds the development of an undcrstanding of false belief. They tested 4-year-old children and concluded that 4-ycar-olds used the concept of surprise as a mental state concept. However, they did not completcly cxplain how or when the development of the concept of surprise comes about.

Recent attempts to answer these concerns come fronl two lines of research: One line is concerned with children's understanding of the emotional consequences of beliel~ and dcsircs (Wellman & Baneriee, 1991), and another is concerned primarily with children's cognitive theories of emotion (Hadwin & Perner, 199 I).

Wellman and Banel~jee's (1991) studies of children's explanations for another's surprise suggested that understanding of cmotions begins with an understanding of the role of desire and is followed by an understanding of the role of beliefs. Their evidence comes from the reports of 3- and 4-year-olds who, in explaining another's surprise, appropriately refer to a protagonist's mistaken beliefs (Harris, 1991), inferring the protagonist's antecedent beliel~ from his consequent emotional state.

Hadwin and Perner's ( 1991 ) study, in contrast, showed a developmental lag in children's understanding that surprise is a function of a match or mismatch between a previously held belief and reality. In their research, children were given stories and asked to rate the protagonist's reaction to the story outcome. The results showed that young children generally interpreted surprise as something pleasant rather than something unexpected.

Thc diffcrcnces between the rcsearch of Wellman and Banerjee (1991) and Hadwin and Perner (19911 concern both the time course lbr understanding surprise as well as the necessary prerequisites for understanding surprise. Whereas the l~rmcr holds that an understanding of surprise comes at about 3 to 4 years of age, the latter view holds that this understanding comes later, at about 5 years of age. And whereas the former holds that t, nderstanding surprise is independent of an understanding of falsc belief, the latter holds that this is a necessary prerequisite to an adult-like understanding of surprise.

Wellman and Gelman (1992), in a review of the theory-of-mind literature, summarized the current status of the controversy:

More recent research on children's understanding of surprise shows that pre- schoolers' organization of emotional understanding encompasscs belief as well as desire states, althougll there is controversy about whether understanding of surprise appears :u about 3 years with the first understanding of belief (Wellman & Baner- jcc, 1991) o." later in the preschool years (Hadwin & Pcrncr 1991). (p. 353)

So the questions t'emain: What do 3-year-olds, who do not understand that others can have false bclicl~, deem as surprising'? And does an understanding of


false belief make an adult-like understanding of surprise possible'? Do children recognize that surprise results both from the unexpected receipt of desirable goods and from the failure of expectation for a desired good'?

A review of Wellman and Banerjee's (1991) methods reveals that Wellman and Banerjee specified subjects' desires and used these specifications to judge happy and surprised. They found that 3-year-olds could not accept that one was both unhappy and surprised. We propose to test this hypothesis using vastly different means. We ask children to judge events as surprising and then infer what principle was important in their natural judgements. Given past research, this principle should be desirability.

In our first attempt to answer these basic questions, a pilot study was conducted using 30 preschool children from the Toronto area. Children were shown objects (such as a jack-in-the-box and a box with a toy inside) and asked to assess the knowledge state of another (a puppet) who had not seen the boxes. We tbund that 3-year-olds and 4-year-olds differed in their assessment of both the knowledge state and the surprise reaction of the other. Younger children unwittingly contradicted themselves by (a) ascribing the knowledge of the content of the box to the other and then (b) stating that the other would be surprised upon encounter- ing this content. Older children consistently stated that if the other did not know the content of the box, he would be surprised, and if the other did know the content of the box, he would not be surprised. Although the results were intrigu- ing, this experiment was confounded in two ways. First, the same "yes" response was incorrect Ior all know questions, and correct across all surprise conditions, hence, a "yes" bias would result in all answers to the know questions being incorrect, and all answers to the surprise questions being correct. Second, in retrospect it was conceivable that jack-in-the-boxes really are surprising (even if one knows their content) simply because they spring up suddenly. Our next experinaent therefore eliminated the link between "'yes" responses and correct answers to the surprise questions and used stimulus objects for which the answer could not be based on a sirnple startle judgement.

The tbllowing general predictions were made about the performance of children on surprise questions: (a) Children will be able to correctly predict another's surprise around age 4; (b) their understanding of surprise will be contingent on their understanding of false belief; only 4-year-old children will understand surprise as the overturning of one belief for another; and (c) children will attribute surprise at an earlier age (i.e., that age reported by Wellman & Bartsch, 1988) when the surprise involves an unexpected treat rather than a failed expectation (trick).

E X P E R I M E N T 1

Method

Subjects. The subjects were 36 boys and 31 girls who were volunteers from various daycare facilities in Metropolitan Toronto, as well as from the visiting


population of the Ontario Science Centre in Toronto, Canada. The children ranged in age from 3;0 to 8;4 years. The 3-year-olds ranged from 3;0 to 3;11 (median = 3;5, N = 20); the 4-year-olds ranged from 4;0 to 4;11 (median = 4;4, N = 16); the 5-year-olds ranged from 5;0 to 5;10 (median = 5;7, N = 13); and the 6÷-year-olds ranged from 6;0 to 8;4 (median = 7;2, N = 18).

Materials. For this experiment a puppet named "Bobby" and a series of surprising objects in containers were used. The surprising events were of two types: The "unexpected treats" items were candy in a toothpaste box and a teddy bear in a plain red and white box. The "failed expectations" items were rocks in a crayon box and rocks in a Smarties (a well known candy) box. These constitute the tricks mentioned in the title of this article.

Unexpected Treats. There were two identical toothpaste boxes; one contained toothpaste and the other contained jellybeans. There were two identical plain red and white boxes; one concealed a small bear figure holding balloons behind the door and the other did not conceal anything.

Failed Expectations. There were two identical crayon boxes; one contained crayons and the other contained rocks. There were two identical Smarties boxes; one contained rocks and the other contained Smarties.

In addition to these stimulus objects, a puppet ("Bobby") was used as the "'other" person whose surprise was to be predicted. Past research has shown (and our experiment confirms) that a puppet works as well as another person in this type of experiment (Sodian & Wimmer, 1987).

Design and Procedure. Before questioning, the experimenter introduced Bobby to the child. The child subject sat at one end of a table, Bobby at the opposite end, and the experimenter between the two. The objects were kept in an open briefcase with the lid blocking the subject's view of the objects inside. Children were told, "We are going to play a game. We know that Bobby is a puppet. But we want you to pretend that Bobby is a person just like you and me, OK'? This means that Bobby can see things just like you and I can see things. And, Bobby can think things just like you and 1 can think things, OK?" In this experiment, the questions concerning the child (called self questions) preceded questions about Bobby (called other questions).

There were four tasks, two involving unexpected treats and two involving failed expectations, each of which took the following form. The order of presentation of these tasks was randomized. The child was introduced to a container while the puppet was absent, and asked by the experimenter, "Do you know what is in the box?" and, "What do you think is in the box?" The experimenter then revealed the unexpected contents to the child, and asked, "Were you surprised when you saw what was inside?"

After the true contents of one of the containers had been revealed to the child, the puppet was introduced and the child was reminded that the puppet had not seen what was in the box. The child was then asked, "Does Bobby know what is

32 Rick Mackaren and David Olson

in the box'?" and, "What does Bobby think is in the box'? Does Bobby think there is toothpaste (the standard content of the box) in the box'? Or does Bobby think there are candies (the nonstandard, but actual content of the box) in the box'?" The correct answers are, "No," and, "Bobby thinks there's toothpaste in the box." These two questions are variants of the standard "false-belief" questions (Wimmer & Perner, 1983),

The box, with its nonstandard contents, was then placed in front of the child along with an identical box containing its usual or standard content (the contents of both were again made known to the child) and the child was asked, "'Which ol ' these two objects will surprise Bobby'?" and, "Can you point to the one whicl~ will surprise Bobby'?" The correct answer was the box with its nonstandard or unexpected contents.

The puppet, Bobby, was then shown the nonstandard, unexpected contents of the box. The order of presentation of standard and nonstandard contents was fixed. Alter the child observed this revelation to the puppet, he or she was asked, "What does Bobby think is inside the box we 've just shown him?" followed by, "Does Bobby think there is toothpaste (the standard content of the box) in the box or does Bobby think there are candies (the nonstandard but now revealed content of the box) in the box'?" The correct answer in this case was, "Candies ."

Table 1. Predicted Response Patterns for All Think and Surprise Question Pairs: (A) Belief-Based Response Patterns Hypothesized for 4-Year-OIds and Older Children and (B) Desire-Based Response Patterns Hypothesized for 3-Year-Olds

False Belief True Belief

Think Surprise Think Surprise

(A) Toothpaste boxes toothpaste candy candy toothpaste

(pass) (pass) (pass) (pass) Unlabeled boxes nothing Teddy Teddy nothing

(pass) (pass) (pass) (pass) Smarties boxes Smarties rocks rocks Smartics

(pass) (pass) (pass) (pass) Crayons boxes Crayons rocks rocks Crayons

(pass) (pass) (pass) (pass) (B) Toothpaste boxes candy candy candy candy

(I;ail) (pass) (pass) (fail) Unlabeled boxes Teddy Teddy Teddy Teddy

(fail) (pass) (pass) (fail) Smarties boxes rocks Smarlies rocks Smarties

(fail) (fail) (pass) (pass) Crayons boxes rocks Crayons rocks Cn, yons

(fail) (fail) (pass) (pass)


For ease of reference this item will be referred to as the second "think" question. Note that unlike the first think question in this case the question is based on a true belief.

The puppet was then taken out of view, and the child was again shown both boxes of the pair, that with the standard content and that with the unexpected content (recall that Bobby had seen only the latter), and was asked, "Will Bobby bc surprised when he sees this, or will Bobby be surprised when he sees this'? Point to the box that will surprise Bobby." The correct answer in this instance was, "toothpaste," because Bobby has just seen and so expects candies. This item is referred to as the second "surprise" question. Table I outlines the predicted patterns of response for younger and older children.

SUMMARY OF QUESTIONS

Self

"Do you know what is in the box'?" "What do you think is in the box'?"

Unexpected contents revealed to child, "Were you surprised when you saw what was inside?"

Other

"Does Bobby know what is in the box'?" "'What does Bobby think is in the box'? Does Bobby think there is toothpaste

(the standard content of the box) in the box'? Or does Bobby think there are candies (the nonstandard, but actual content of the box) in the box'?" (First "think" question; false belief)

"Will Bobby be surprised when we show him candies or toothpaste'?" and, "Can you point to the one that will surprise Bobby'?" (the contents of the box are now revealed to the child and the puppet). (First "surprise" question)

"What does Bobby think is in the box'? Does Bobby think there is toothpaste in the box'? Or does Bobby think there are candies in the box?" (Second "'think" question; true belief)

"Will Bobby be surprised when we show him candies or toothpaste'?" and, "Can you point to the one that will surprise Bobby'?" (Second "surprise" question)

RESULTS

Questions for Self For each of the "self" questions children were assigned a composite score for the four items. For example, for the "know" question, when asked if the child knew


Table 2. Number of Children at Each Age Failing or Passing Questions About What the Self Knows, Thinks, or is Surprised By

3-Year -Olds 4 -Year -Olds 5-Year -Olds 6 +-Year -Olds

Know

Fail I 2 I 3 Pass 19 14 12 15 Think--Failed expectations and unexpected treats Fail 0 0 0 0 Pass 20 16 13 18

Surprise--Failed expectations and unexpected treats Fail 0 0 2 I Pass 20 16 1 I 17

Note. "'Fail" refers to scores of 0 to 2 and "Pass" refers to those of 3 to 4 correct out o1' 4.

the contents of the box, the expected answer was "yes" Ik~r the Smarties item because the box labeling depicted Smarties. Subject scores for this item thus ranged fi'om zero (0) "'yes" responses out of four items to four (4) "yes" responses out of four items. Similarly, for the think question, the expected answer was "Smarties" for the Smarties item because the colorful labeling was good reason for the child to think Smarties were in the box. Finally, for the surprise question, the expected answer for all items was "'yes" as the box contents always violated expectations. Responses were then collapsed into two groups: scores from 0 to 2 expected were taken as failing the question and scores of 3 to 4 were taken as passing the question.

Know Children's responses to the self-know questions (e.g., "Do you know what is in the box'?") were compared across each of the four age groups to determine if performance on the know question improves with age.

The scores for each age group did not differ markedly, and as Table 2 shows, the majority of children, regardless of age, appropriately believed they knew the contents of the box.

Think Children's responses to the self-think question (e.g., "What do you think is in the box'?") are also shown in Table 2. As expected, the majority of children, regardless of age, based their guess as to the actual contents of the boxes on the basis of the container by, for example, saying Smarties would be inside the Smarties box.

Surprise Children made judgements as to their own experience of surprise for each of the four items. Performance on this surprise question showed most subjects reported


having experienced surprise regardless of the desirability of the contents of the boxes.

Questions for the Other The distributions of responses across the various questions are such that children tend to either pass the majority of items on a given question or to fail them. The overwhelming majority of 3-year-olds get 0 or 1 of 4 items of a given question type correct whereas 5-year-olds get 3 or more of the 4 items of a type correct.

The distribution of responses for 4-year-olds is not uniform across the know, think and surprise questions. For the know question, the distribution of responses for the 16 lbur-year-olds is negatively skewed (3 scored 0 items correct, 2 scored 2 correct, 3 scored 3 correct, and 8 scored 4 correct). For the think question, the distribution of responses lor 4-year-olds is bimodal (3 scored 0 correct, 3 scored I correct, I scored 2 correct, 4 scored 3 correct, and 5 scored 4 correct). Finally, for the surprise questions, the responses of 4-year-olds were normally distributed ( 1 scored 0 correct, 4 scored 1 correct, 6 scored 2 correct, 3 scored 3 correct, and 2 scored 4 correct). Thus, many 4-year-olds scored in the middle range. How- ever, when these items were differentiated by type children tended to have succeeded or failed on that item. For the failed expectation items the distribution of responses is positively skewed (7 scored 0 correct out of 2, 6 scored 1 correct, and 3 scored 2 correct) whereas for the unexpected treats items the distribution of responses was negatively skewed (3 scored 0 correct, 5 scored 1 correct, and 8 scored 2 correct), Thus, on failed expectations items 4-year-olds were more like 3-year-olds, and on unexpected treats they were more like 5-year-olds.

Scoring For each of the questions for the other, children were assigned a composite score for stimuli ranging from no correct responses out of four stimuli (0) to four correct responses out of four stimuli (4). Subject responses were then collapsed into two groups; scores from 0 to 2 correct were taken as failing the question and scores of 3 to 4 were taken as passing the question. This classification had the additional advantage that it met the requirement for minimum expected cell frequencies. The same criteria were consistently employed across all statistical tests (except where alternative criteria were more informative).

Know Children's responses to the know questions (e.g., "Does Bobby know what is in the box'?") were compared across each of the four age groups to determine if performance on the know question improves with age.

The scores for each age group differed markedly, and as Table 3 shows, the major shift occurs between 3 and 4 years of age, ×"(I, N = 36) = 19.8, p < 0.0001.


Table 3. Number of Children at Each Age Failing or Passing on Questions About What the "Other" Knows, Thinks, or is Surprised By

3-Year-Olds 4-Year-Olds 5-Year-Olds 6 +-Year-Olds

Know Fail 20 5 3 2 Pass 0 I 1 10 16 Think--Failed expectations Fail 16 6 I I Pass 4 10 12 17 Think--Unexpected treats Fail 17 5 1 1 Pass 3 I 1 12 17 Surprise--Failed expectations Fail 19 10 5 0 Pass I 6 8 18 Surprise--Unexpected treats Fail 13 6 5 1 Pass 7 I 0 8 17

Note. "Fail" refers to scores of 0 to 2; "Pass" refers to scores of 3 to 4.

Think Child responses to the think quest ions (e .g . , "Wha t does Bobby think is in the

box'?") are also shown in Table 3 for each type of item, tailed expectat ion or

unexpected treats.

Unlike 4-year-old and older children, the majority of 3-year-olds were unsuc-

cessful on the first think quest ion for both the tailed expectat ion items X2(I, N =

36) = 6.8, p < 0.01 and the unexpected treats x2( l , N = 36) = 10.8, p < 0 .001,

c la iming Bobby would think the box contained the actual contents rather than the

depicted contents; most 4-year-olds and older children were successful. By in-

spection, it is obvious that the results for failed expectat ions items mirror the

unexpected treats i tems.

There was a significant cont ingency between the answers g iven to the know

and think questions. Table 4A shows that those who incorrectly said the other

Table 4. Contingency Tables Showing the Number of Subjects Who Pass/Fail the Know, Think, and Surprise Questions

A. Know B. Surprise

Fail Pass Fail Pass

Think Fail 21 0 25 3 Pass 9 37 8 31


knows also incorrectly said the other would think the box contained the actual contents, X2(I, N = 67) = 34.5, p < 0.0001. Consequently, success on either or both the know or tbink questions indicates that the majority of children come to understand belief at 4 years of age. Performance on both these items suggest that the majority of children first come to understand false belief at about 4 years of age.

Surprise The first comparison is made by combining the two types of surprise items. Children made 8 judgements as to what would surprise Bobby, 2 for each item, yielding a composite surprise score between 0 and 8. Scores between 0 and 4 were taken as failing the surprise question and scores from 5 to 8 were taken as passing. Performance on the surprise question shows a significant increase across the four age groups, as shown in Figure 1, X2(3, N = 67) = 8.8, p < 0.01. The mean percentage of 3-year-olds passing the surprise items (40% passing) was well below the percentage of older children passing (75% passing), with the difference between 3- and 4-year-olds being significant, X2(3, N = 36) = 4.4, p < 0.05.

When the items (e.g., failed expectations and unexpected treats) are dis- tinguished, important differences appear. Children's responses to the surprise question (e,g., "'Which one will surprise Bobby'?") are also shown in Table 3.

For the failed expectations items, 5% of 3-year-olds and 38% of 4-year-olds scored 3 or 4 items correct. This result stands in contrast to the score of 5- and 6+-year-olds, 62% and 100% of whom scored 3 or 4 items correct, X:'(3, N = 67) = 36, p < 0.001. For unexpected treats items, 35% of 3-year-olds; 62% of 4- year-olds; 62% of 5-year-olds, and 94% of 6 +-year-olds scored between 3 to 4 items correct. Again, these differences are significant, ×2(3, N = 67) = 14.3, p < 0.01.

P e r c e n t

P a s s i n g

Figure 1.

100

80

60

40

20

0

j i i i i I

3 4 5 6

Age

Percentage of children passing surprise questions.


Figure 2. items.

Percent Passing

lOO

, o

60

40 Tricks

Treats

20

0 3 4 5 6

Age

Percentage of children passing surprise questions for tricks and treats

Recall that younger children were predicted to base their judgements on the desirability of the content (an object-oriented theory of surprise) whereas the older children were predicted to base their judgements on the overturning of a prior belief of the other person (the mental-state theory of surprise). To examine this possibility, the patterns of response for the items involving desirable surprises (treats) were separated from those involving undesirable surprises (tricks). One of the surprise questions for each item required the child to choose a desirable object as the surprising object and one required the choice of the undesirable item as the surprising item. Responses to these questions are shown in Figure 2. This figure indicates that there were noteworthy differences between the tasks requiring desirable and those requiring the choice of undesirable objects as surprising. Change over the age range studied is manifest only for the undesirable surprise, the trick item. In order to exanaine the undesirable surprises in more detail, it is necessary to consider the first surprise question for one item type and the second surprise question for the other item type. Hence, for the unexpected treats (e.g., the candies in toothpaste box and teddybear in unlabeled box) responses were analyzed for the second surprise question only. Recall that in this case children had just witnessed the puppet observe the actual desirable content of the box, and were now faced with two boxes, one with the attractive but now known contents and the other with unattractive but unknown contents. If children base their responses on the other's mental state, they will now predict that the other will judge the containers with "toothpaste" and "nothing" as surprising. If they base their answers on the desirable properties of the contents they will judge the containers with the "candy" and "teddy bear," as surprising.

Toothpaste/Candy: There were 10 candy responses among 3-year-olds, 4 candy responses among 4-year-olds, 3 candy responses among 5-year-olds, and 0 candy responses among 6+-year-olds, ×2(3, N = 67) = 12.6, p < 0.01. Ted- dybear/Empty: There were 11 bear responses among 3-year-olds, 7 bear responses among 4-year-olds, 3 bear responses among 5-year-olds, and 0 bear responses anmng 6+-year-olds, X2(3, N = 67) = 14.9, p < 0.002.


lOO

8o

Percent 60 Passing

40

20

Figure 3.

Toothpaste/candy

Empty/teddy

Crayons/rocks

Smarties/rocks

0 i i i i I

3 4 5 6 Age

Percentage of children passing surprise questions by item.

The failed expectations objects (e.g., the rocks in the crayons box and rocks in the Smarties box) were analyzed for the first surprise question only, because it was t'or the first surprise question that mental state and object-oriented hypoth- eses led to different predictions. If subjects base their responses on the other's mental state, they will predict the other will find "rocks" surprising. If subjects base their answers on the objectively desirable properties of the contents, they will predict the other will find crayons and Smarties surprising.

Crayons/Rocks: There were 19 crayons responses among 3-year-olds, 7 crayons responses among 4-year-olds, 4 crayons responses among 5-year-olds, and 0 crayons responses anmng 6 +-year-olds, X2(3, N = 67) = 36.0, p < 0.001. Smarties/Rocks: There were 17 Smarties responses among 3-year-olds, 13 Smar- ties responses among 4-year-olds, 6 Smarties responses among 5-year-olds, and 2 Smarties responses among 6+-year-olds, X2(3, N = 67) = 33.1, p < 0.001. Thus, there were noteworthy differences between the types of tasks; whereas most 3-year-olds judged the desirable but expected object as surprising, by age 5 years the majority judged the unexpected as surprising whether desirable or not. These results are summarized in Figure 3.

Because the performance on the second surprise question for all items was premised on a true rather than a false belief, a comparison of the first and second surprise questions for 3- and 4-year-olds was warranted. The result of a MAN- OVA showed significant age, F(I , 34) = 8.5, p < 0.01, and trial, F(I , 34) = 9.9, p < 0.005, main effects. The analysis showed no age x trial interaction. Children performed better on the second trial. Thus, although children's performance may be positively affected when explicitly given the other's belief, younger children benefit from it to no greater extent than older children.

The special difficulty 3-year-old children had in acknowledging that an unexpected but undesirable item could be surprising is further indicated by comparing performance on the think and surprise questions for every item. Table 5 shows how virtually all of the items in Column 1, which involve the puppet's false


Table 5. Actual Response Patterns for all Think and Surprise Question Pairs Showing Belief and Desire-Based Response Patterns

T h i n k - S u r p r i s e

False Belief True Belief

I t em Age N Belief Desire Belief Desire

Toothpaste 3 2(/ 0 16 8 I 0 4 16 8 5 II 3

5 13 10 I 10 2

6 + 18 17 0 18 0

Unlabeled 3 20 I 14 8 I I 4 16 8 6 9 7

5 13 7 2 I0 2 6 + 18 14 0 18 0

Smarties 3 20 0 12 16 16 4 16 3 6 12 12

5 13 7 1 12 12

6 ~ 18 17 I 17 17

Crayons 3 20 0 16 12 12 4 16 6 3 12 12 5 13 9 2 12 12

6 + 18 17 0 18 18

belief, were failed by younger children and passed by older children. For the youngest children, when items involved unexpected treats (a teddy in a box vs. an empty box; candy in a toothpaste box) children claimed both that the other would think that the box had its actual contents and that the other would be surprised by those contents. When rocks were the surprising content they were much less likely to claim that the other would think "rocks" yet be surprised by rocks. In a word, the youngest children thought "Teddies" and candies surprising but not rocks. This analysis goes some distance in correcting the impression given by Figure 2, namely, that even the youngest children could pick the surprising object when that surprising object was a treat. The combined " think- surprise" scores show that in many cases the youngest children are claiming both that the subject knows and yet will be surprised. Table 5 shows that this was particularly common for 3-year-olds on items in which the surprise was a treat.

Finally, consider the relation between performance on the false-belief (think) question and judgements of surprise. If understanding false belief is a necessary condition for making correct surprise judgements, then we expect that children who succeed on the one will succeed on the other.

In order to assess this possibility, a contingency relation was calculated for the two tasks. The results are shown in Table 4b, which shows clearly that those who pass false belief are also able to make systematic surprise judgements; those who fail the think question also fail the surprise question, ×2(i, N = 67) = 28.2, p < 0.0001.


DISCUSSION

Results from this study confirm one of Wellman and Bartsch's (1988) findings: Four-year-olds understand that surprise results when one gets what is not expected. It goes beyond that study in three ways, first, to show that a mature understanding of surprise depends on an understanding of false belief and second, in showing that even 3-year-olds have a concept of surprise, albeit somewhat different from that held by 4-year-olds. Finally, it shows that children have particular difficulty in recognizing tricks as surprises.

The results confirmed our expectation that in their 4th and 5th year children come to understand that surprise is based on the violation of one's beliefs. Consider first their understanding of false belief. Children's responses to the think and know questions improved markedly between the ages of 3 and 5 years. The younger children said that the other knew (when he did not) and would think the box contained its actual contents rather than the contents depicted on the box, the contents which moments earlier they had indicated they thought to be in the box. The majority of the 4-, 5-, and 6 +-year-old children said the other would think the box contained the objects depicted on the box. These children were taken as understanding false beliefs. These results for the think question were consistent with the results obtained by Wimmer and Perner (1983) in their false- belief task.

How is false belief implicated in children's understanding of the concept of surprise'? Three-year-olds' performance on the surprise question paralleled their pertbrmance on the think questions. Typically they say the other will think the true content of the box is in the box (ignoring the misleading label and picture on it) and yet that person will be surprised at its content. Five-year-olds, on the other hand, nearly 70% of the time say the other will think the box has the content depicted on the label and will be surprised at the content when it deviates from expectation. By 6 + years of age there were few errors. The contingency between false belief and surprise demonstrates that age alone cannot predict performance on the surprise task whereas performance on the false-belief task can. Under- standing of false belief, we conclude, is necessary for understanding surprise.

However, that is not to say that 3-year-olds have no ur.derstanding of surprise. As long as the surprising contents are desirable, 3-year-olds perform almost as well as 6+-year-olds. Their limited understanding appears most clearly when desirability conflicts with belief violation. On these tasks, 3-year-olds' performance falls as low as 5% and at its highest reaches 50%. Thus Wellman and Bartsch's conclusion that children understand surprise by 4 years of age must be qualified by distinguishing understanding of surprise, which is independent from an understanding of false belief and is indicated in this study by the 3-year-olds' selection of the surprising item as long as it was desirable, from the more sophisticated understanding of surprise, which is contingent upon an understanding of false belief and was achieved by about half of the 4-year-olds and the majority of 5- and 6 +-year-olds whether the item was desirable or not. Only


these older children could acknowledge that finding rocks in a candy box could be surprising. These findings therefore compliment Wellman and Banerjee's (1991) claim that 3-year-olds could not find objects both surprising and undesirable.

It could be argued that these results indicate a change in children's understanding of the word "'surprise" as opposed to increased understanding of the concept of surprise. We acknowledge that it is virtually impossible, at least with existing methodology, to distinguish the two. The evidence reported above indicates both that the children are familiar with the term and that they consistently apply it to the unexpected treats when they are 3-years-old. If young children were confused about the term they would apply the term randomly to any item. Instead, they apply the term surprise unitbrmly to treats. So the very young children know the term. But it is reasonable to infer the 3-year-old may mean something different by the term than the older children. This interpretation is supported by the fact that their understanding of surprise in this study was contingent upon their understanding of false belief. With the onset of an understanding of false belief they apply the term to a different range of items. Their familiarity with the word would not account for this contingency.

Notice that our claim is not that children experience surprise only at age 3 or 4: rather it is that they understand surprise, an understanding they can use to ascribe that state to others. To make such an attribution requires that the child have the concept of surprise. It is that concept, we have suggested, that is examined in this study.

Nonetheless, it must be acknowledged that the question in the current experiment always employed the term "surprise." In an attempt to extract the child's ability to ascribe surprise to others, without asking which object would surprise Bobby, we conducted a replication in which we did our utmost to get younger children to understand what we meant by the word "surprise" by identifying it with a distinctive facial expression. If they simply failed to grasp our term but had the concept of surprise in place, it is possible that the facial expression would clarify for them what we wanted. If they lack the concept of surprise, as we have suggested, a pictured facial expression is unlikely to communicate it to them any better than the term itself would.

EXPERIMENT 2

In this experiment, rather than merely asking which object would surprise the "other," children were asked to select a pictured facial expression of surprise to pair with the surprising event. The purpose of the study was to attempt to replicate the finding of Experiment 1 using an alternative response procedure, namely, choosing a pictured facial expression appropriate to surprising and non- surprising events. As mentioned, the choice of a pictured representation of surprise was based on the following considerations. If young children are simply


unfamiliar with the term surprise, and they in fact have an understanding of the concept of surprise which entails the unexpected, then we may expect children under 5 years of age will successfully select the appropriate facial expression. But if young children lack an understanding of the concept of surprise as depend- ing upon the mental state of another, they will have no more basis tbr systemat- ically assigning the surprised facial expression than they would for assigning the verbal label.

Children 3 to 6 + years of age were asked to match two faces, one surprised and one not-surprised, to two boxes with differing contents. Only one of the tasks from the previous experiment, specifically the failed expectation task involving Smarties and rocks in Smarties boxes, was used in this experiment. Recall that in Experiment 1 this failed-expectations task was one of the more difficult of the tasks used. If children understand that the puppet expects Smarties to be in both of the boxes, children should, if they understand that surprise depends on the overturning of one belief for another, match the surprised lace to the Smarties box containing rocks.

Method

Subjects. The subjects were 37 boys and 41 girls who were volunteers from the visiting population of the Ontario Science Centre in Toronto, Canada. The children ranged in age from 3:0 to 7;1 years. The 3-year-olds ranged from 3;0 to 3;11 (median = 3;5, N = 19); the 4-year-olds ranged from 4;0 to 4;11 (median = 4;6, N = 23); the 5-year-olds ranged from 5;0 to 5;11 (median = 5;6, N = 19); and the 6+-year-olds ranged from 6;0 to 7:1 (median = 6;8, N = 17).

Materials. The task employed was one of the failed-expectations tasks used in the previous experiment. Two Smarties boxes were used: one contained the standard contents (e.g., Smarties) and one contained rocks.

In addition to these objects, a character ("Fred") was used as the other in this experiment. Fred's face was line drawn on two 6" x 6" cards, one of which showed a surprised face (with raised eyebrows and rounded mouth) and one a not-surprised facial expression.

Design and Procedure. Children were first introduced to Fred's not-surprised face. They were told, "This person here, named Fred, is a friend of mine. Right now he is not happy and not sad. We will put Fred away now and I'll show you something. But we won' t show Fred. He'll be asleep. OK?"

Children were then introduced to two boxes. The box with unexpected contents was placed in front of a child along with an identical box containing its usual or standard content and the child was told, "'Here are two boxes- -both are Smarties boxes. Fred thinks there are Smarties inside both boxes. But I'll show you (but not Fred) that one of the boxes has Smarties inside and one has rocks inside. See?"


Children were then introduced to Fred's surprised face. They were told, "This is Fred when Fred is surprised--he's not happy, and he's not sad, he's just surprised. Now, show me which box will make Fred's face do this." The correct answer required them to pair the surprised face to the box with unexpected content, namely, the box with rocks inside.

RESULTS

Fifteen of the 3-year-olds (80%) and 18 of the 4-year-olds (70%) paired the surprised face with the box containing the Smarties whereas only 6 of the 5-year- olds (32%) and 5 of the 6 ÷-year-olds (29%) did so; the difference was significant, X2(3, 78) = 17.6, p < 0.001. Clearly then, most 3- and 4-year-old children think Fred will be surprised upon finding Smarties inside the Smarties box, whereas most 5 and 6 +-year-olds think Fred will be surprised finding an unexpected content, rocks.

DISCUSSION

The results of Experiment 2 are largely congruent with those obtained in the comparable task in Experiment l. In particular, children tend to deem surprise to be dependent upon the unexpected by 5 years of age. Although it may appear that a majority of 4-year-olds pass the surprise item when expressed verbally in Experiment 1, whereas a majority of 4-year-olds failed in Experiment 2, in fact the performance was similar on the comparable items. In Experiment 1,6 out of 16 four-year-olds got both failed-expectations items correct; 7 out of 23 got the comparable item correct in Experiment 2. Hence, we can infer that having to answer a question containing the word surprise (as in Experiment I) does not account for the 4-year-olds' difficulty. Combining the results of the two studies, approximately 36% of 4-year-olds got all such items correct, whereas 65% of 5- year-olds did. The 3- and 4-year-olds make errors in judging the surprise reac- tions of the character about 80% of the time, whereas 5- and 6 +-year-olds make errors less than 35% of the time.

In general, the results of Experiment 2 served to uphold the findings of Experiment I. But in addition, Experiment 2 resolved two important methodological issues. Because there was only one surprise question posed to each child, and because each child made judgements for only one stimulus item, the earlier concerns about repeated questioning and fixed order of presentation of stimuli are alleviated. The performance of children was largely unaffected by the change in methodology.

Experiment 2 also served to resolve an important theoretical issue. Recall that children in this experiment were asked to match a face to a Smarties box, and it was therefore unnecessary for children to respond to a question containing the


word surprise. This experiment required only that the child recognize the facial reaction of Fred when he sees an unsurprising versus surprising event.

The results of this experiment therefore lend strong support to the hypothesis that children acquire a concept of surprise as involving the overturning of belief at about 4 or 5 years of age.

GENERAL DISCUSSION

Children's understanding of the concept of surprise undergoes an important alteration between 3 and 5 years of age. Three- and 4-year-olds believe that another will be surprised if he or she is presented with a subjectively valued object such as candies or a teddy bear regardless of the other's prior mental state. At 5 years of age they recognize that another's surprise will depend not on the qualities of the object but upon the other's prior expectancies or mental state independently of the intrinsic features of the surprising object. For this reason, their understanding of the concept of surprise is contingent upon their understanding of another's false belief.

How can we characterize the 3-year-olds' concept of surprise'? Their judgment of surprise showed a large degree of stimulus dependence; when a pleasing object was pitted against an unexpected one, the 3-year-old chooses the pleasing one, suggesting that the child simply chooses those which are in accord with the puppet's desires. Yet the data are insufficient to support this conclusion, as we have no evidence to indicate that children judged the character to desire or "want" any of the "surprising" objects. We infer that young children are taking surprise as a state that follows the presentation of objects with particular desirable properties. The conclusion on this point is similar to that of Wellman and Banerjee (1991) who found that 3-year-olds could not accept that one was both unhappy and surprised. We arrive at a similar conclusion by quite different means. We asked children to judge events as surprising and then inferred what principle was important in their natural judgements. We found that the principle was desirability of the surprising item. Whereas a majority of 4-year-olds readily recognized that an unexpected treat would be a surprise for the puppet, they failed to recognize that surprise would be produced by failed expectations--the simple overturning of one belief for another.

An understanding of surprise requires the understanding that others may have beliefs and expectations that may or may not be fulfilled. An adult-like understanding of surprise, that is, the ability to recognize that even undesirable items may be surprising, depends on an understanding of false belief, as Experiment 1 showed. On this point our results are not inconsistent with those reported by Hadwin and Perner (1991). Understanding surprise, we may say, is simply another manifestation of children's more general understanding of these mental states.


REFERENCES

Astington, J.W., Harris, P.L., & Olson, D.R. (1988). Introduction. In J.W. Astington. P.L. Harris. & D.R. Olson (Eds.), Dew'loping theories qf mind (pp. 1-15). Cambridge: Cambridge University Press.

Hadwin, J., & Pemer, J. (1991). Ple~,sed and surprised: Children's cognitivc theory of emotion. British Journal ,~" Develol,mental Psychology, 9, 215-234.

Harris. P.L. (1991). Preface. British Jourmd ~y'Developmental Psychology. 9. 190. Hogrefe, J., Wimmer, H., & Perner, J. (1986). Ignorance versus fi, lse belief: A developmental lag in

attribution of epistemic states. Child Det'elOl~ment. 57. 567-582. Perner, J. (1988). Developing semantics for theories of mind: From propositional attitudes to rnental

representation. In J.W. Astington, P.L. Harris, & D.R. Olson (Eds.), Developing theories of miml ( pp. 14 I - 172). Cambridge: Cambridge University Press.

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

Sodian, B., & Wimmer, H. (1987). Children's understanding of inference as a source of knowledge. Child Development. 58, 424-433.

Wellman, H., & Banerjee, M. (1991). Mind and emotion: Children's understanding of the emotional consequences of beliefs and desires. British Joto'tlal of Developmental Psychology. 9. 191- 214.

Wellman, H., & Bartsch, K. (1988). Young children's reasoning about beliefs. Cognition, 30, 239- 277.

Wellman, H.. & Gelman, S. (1992). Cognitive development: Foundational theories of core domains. Annual Revien, of Psychology. 43. 337-375.

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

Trick or treat: Children's understanding of surprise

Documents