Child Development, July/August 1999, Volume 70, Number 4, Pages 853865

Theory of Mind, Emotion Understanding, Language, and Family Background: Individual Differences and Interrelations
Alexandra L. Cutting and Judy Dunn

Individual differences in young childrens social cognition were examined in 128 urban preschoolers from a wide range of backgrounds. Comprehensive assessments were made of childrens false-belief understanding, emotion understanding, language abilities, and family background information was collected via parent interview. Individual differences in childrens understanding of false-belief and emotion were associated with differences in language ability and with certain aspects of family background, in particular, parental occupational class and mothers education. The number of siblings that children had did not relate to their social cognition. Individual differences in false-belief and emotion understanding were correlated, but these domains did not contribute to each other independently of age, language ability, and family background. In fact, variance in family background only contributed uniquely to false-belief understanding. The results suggest that family background has a signicant impact on the development of theory of mind. The ndings also suggest that understanding of false-belief and understanding of emotion may be distinct aspects of social cognition in young children.

INTRODUCTION It is now widely accepted that by 4 years of age most normally developing children have acquired an understanding of mind. That is, they are able to understand a persons actions in terms of that persons desires, thoughts, beliefs (which may be true or false), and emotions (for reviews, see Astington, 1993; Mitchell, 1997). Much of the work in this area has been concerned with normative development and group means, focusing on the age at which children develop a theory of mind, and the conditions under which they are able to display it (e.g., Freeman, Lewis, & Doherty, 1991; Moses & Flavell, 1990; Sullivan & Winner, 1993). Although this has resulted in a fairly clear understanding of what constitutes the normal development of social cognition, two areas within the eld remain neglectedindividual differences and the effects of family background. Individual Differences in Social Cognition It is only in recent years that researchers have begun to focus on individual differences in childrens understanding of belief and emotion, and the factors associated with these differences (Dunn, Brown, Slomkowski, Tesla, & Youngblade, 1991; Jenkins & Astington, 1996). Yet an understanding of individual differences is essential if we are to discover the factors that inuence development, and ultimately the mechanisms involved (Astington, 1993; Bartsch & Estes, 1996). Although few studies have specically addressed the issue of individual differences, several factors that

inuence theory of mind development have been identied. Two areas that have received particular attention are the inuences of family environment and of language. In one of the rst studies to examine individual differences in social understanding, Dunn et al. (1991) observed children aged 33 months at home with their mother and siblings. The children then were tested for both false-belief understanding and emotion understanding (specically, their ability to make inferences about the relations between particular situations and basic emotions) at 40 months. Clear associations were found between family measures at 33 months (such as family talk about feelings, and cooperation between siblings) and both false-belief and emotion understanding at 40 months. In addition, individual differences in social pretend play with mothers and siblings at 33 months related to social understanding at 40 months (Youngblade & Dunn, 1995). Further support for the importance of family environment comes from recent studies that have found relations between performance on false-belief tasks and the size of a childs family, and in particular the number of siblings that a child has (Jenkins & Astington, 1996; Lewis, Freeman, Kyriakidou, Maridaki-Kassotaki, & Berridge, 1996; Perner, Ruffman, & Leekam, 1994). A second area that has received attention in examining individual differences in theory of mind is that of language. In a key paper, Happ (1995) showed
1999 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/99/7004-0004

854

Child Development

that verbal ability, as measured by the British Picture Vocabulary Scale (BPVS, a test of receptive vocabulary), was clearly linked to ability to pass false-belief tasks, both in normally developing children and in children with autism. Jenkins and Astington (1996) replicated this effect using the Test of Early Language Development (Hresko, Reid, & Hammill, 1981), a measure of overall language development, both expressive and receptive), nding a strong relation between general language ability and false-belief performance. Childrens ability to pass a false-belief task also may depend upon their comprehension of the narrative involved (Lewis, Freeman, Hagestadt, & Douglas, 1994). In addition to childrens own language ability, the experience children have in participating in discourse with their families has been related to their theory of mind development (Dunn et al., 1991; Welch-Ross, 1997). These experiments highlight the fact that language and theory of mind are intimately related, but several questions remain to be answered. In particular, it is not known whether certain aspects of language are particularly important for theory of mind development, although arguments have been put forward emphasizing the role of almost every aspect of languagefrom syntax (e.g., de Villiers, 1995; Feldman, 1988) to semantics (e.g., Moore, Pure, & Furrow, 1990; Olson, 1988) and pragmatics (Baron-Cohen, 1988; Eisenmajer & Prior, 1991). The issue of cause and effect also is unclear; language may inuence theory of mind development, the opposite may be true, or they may inuence each other. It also has been argued that many theory of mind tests are linguistically complex, which may in part account for correlations between language ability and test performance (e.g., Lewis & Osborne, 1990). Correlations between language ability and theory of mind performance have now been established, however, for a wide range of theory of mind measures, some less verbal than others, so the linguistic demands of standard tasks are unlikely to account fully for correlations between language ability and performance on theory of mind tests. Individual differences in emotion understanding have been somewhat more researched than individual differences in false-belief understanding. Individual differences in childrens understanding of emotion, as indexed by their talk about emotion, have been documented in children as young as 24 months (Dunn, Bretherton, & Munn, 1987). Individual differences also have been identied in a variety of related areas, including response to distress (Cummings, Hollenbeck, Ianotti, Radke-Yarrow, & Zahn-Waxler, 1986), comforting (Murphy, 1937), and understanding of the causes of emotions (Denham, 1986) and also are

known to be stable over time (Brown & Dunn, 1996). As with false-belief understanding, individual differences in emotion understanding have been related to family environment, and also possibly to gender (see Harris, 1994, for a review). In particular, family discourse and the quality of family interactions seem to be important (Dunn et al., 1991). The relation between understanding of false-belief and emotion understanding is surprisingly unclear, despite the importance of both domains to social interaction. Little is known, for example, about whether children who have a strong grasp of mental states also are advanced in their understanding of emotions. Nor is it known how, or indeed if, these domains inuence each other in their development. Does seeing another child show an emotional response to a situation that differs from his or her own response help a child understand that people may have different points of view, for example? Evidence that does exist concerning the relation between understanding of false-belief and emotion is mixed. Dunn et al. (1991) found that family measures at 33 months were more strongly associated with emotion understanding at 40 months than with falsebelief understanding, leading the authors to speculate that understanding emotion and understanding belief may develop along different lines, with emotion understanding perhaps being more inuenced by patterns of family interaction than is understanding of belief. In addition, Dunn et al. noted that performance on their emotion understanding and falsebelief tasks did not correlate signicantly. The childrens understanding of false-belief was assessed when they were very young (only 40 months), however, and was limited in scope, focusing only on the ability to explain a characters action in terms of a false-belief. Stronger evidence for a dissociation between understanding of belief and emotion comes from the nding that individual differences in each domain at 40 months had distinct sequelae at the end of the childrens kindergarten year of school (Dunn, 1995). In contrast to these ndings, there is some evidence for a link between understanding of false-belief and emotion. Several studies have shown that, by around 4 years of age, children can appreciate that the emotion a person feels about a given event depends upon that persons perception of the event and their beliefs and desires about it. For example, if an event differs from the persons belief about what will happen, they are likely to feel surprise; they will be happy if they (falsely) believe that they will get what they want (Baron-Cohen, 1991; Hadwin & Perner, 1991; Harris, Johnson, Hutton, Andrews, & Cooke, 1989;

Cutting and Dunn

855

Wellman & Banerjee, 1991). These studies, however, have focused primarily on group means and age norms, rather than investigating individual differences. One study has looked at individual differences in understanding of false-belief and emotion. Hughes and Dunn (1998) conducted a longitudinal study of 50 children aged 47 months to 60 months, examining developmental changes in understanding of falsebelief and emotion, and mental-state conversation with friends. They found that individual differences in understanding of both false-belief and emotion were stable over this time period and were signicantly related to each other. In addition, these differences were reected in childrens use of mental-state terms in everyday conversations with their friends. Family Background and Social Cognition The majority of published studies of theory of mind development have investigated the abilities of middleand upper-middle-class children from generally well-educated families. It may not be appropriate to generalize results from these groups to all children, however, particularly those from impoverished backgrounds. Signicant differences between middle-class and working-class children have been identied in other areas of cognition, such as conservation (Roazzi & Bryant, 1997) and verbal development (Hart & Risley, 1995). Given the apparent importance of aspects of the family environment such as sibling interactions and family discourse about inner states, it may be expected that socioeconomic status (SES) also has an impact on social cognition. The role of family background in the development of childrens understanding of belief and emotion, however, has not been systematically investigated. There is some evidence that family background plays a role in theory of mind development. Children from a Head Start population have been reported to perform poorly on false-belief tests, although in this case, performance was compared only to levels typically reported in the literature (Holmes, Black, & Miller, 1996). Cross-cultural studies also have suggested that family background affects theory of mind development; for example, afuent Indian children outperformed their deprived counterparts in making mental-real distinctions (Wahi & Johri, 1994). In addition, parenting style (authoritarian versus authoritative) may relate to young childrens theory of mind performance (Vinden, 1997), and parenting style is known to vary considerably across different family backgrounds (for a review, see Hoff-Ginsberg & Tardif, 1995). Work on emotion understanding has included

studies of low-income families (e.g., Garner, Jones, Gaddy, & Rennie, 1997) and cross-cultural comparisons (see Saarni & Harris, 1991), which, taken together, hint at poorer emotion understanding in children from deprived backgrounds. Young childrens emotion understanding also has been found to correlate signicantly with their fathers occupational prestige (Dunn et al., 1991), level of education, and occupational class (Dunn & Brown, 1994). In this study, we examine social cognition in a large sample of urban preschoolers from a wide range of backgrounds. The goals of the study are threefold. The rst goal is to examine individual differences in childrens understanding of false-belief and emotion using a comprehensive multimeasure approach to assessment. This is likely to give an accurate assessment of a childs ability, and is more sensitive to individual differences than are single-test approaches, which simply categorize children as passers or failers. The second goal of the study is to investigate links between individual differences in social cognition and individual differences in family background. The third goal is to examine how different aspects of social cognition relate to one another, focusing on understanding of false-belief, emotion understanding, and language ability.

METHOD Participants The participants were 128 children (63 girls, 65 boys) with a mean age of 4.16 years (range 3.49 4.80). The children attended nursery schools in South London and were taking part in a longitudinal study of friendship in young children. The children came about equally from working-class and middle-class backgrounds (see below for details). Sixty-eight percent were White, 26% were Black or mixed race, and 6% were from other ethnic groups. Procedure Each child was seen twice in nursery school (visits conducted, on average, 1 week apart) and once at home (on average, 3 weeks later). A battery of falsebelief, language, and emotion understanding tests was administered over the course of these visits. Family background information was collected from parents. The majority of children participated in all the tests involved in the study. Some did not complete every task, for a variety of reasons (e.g., noncompliance, absence from school), so specic Ns are given for each task with the results.

856

Child Development

False-Belief Tasks A total of eight false-belief tasks were administered, all of which involved predicting, explaining, or recalling a false-belief. Three types of task were used. 1. Unexpected-location tasks. Four unexpected-location stories were used, based on the tasks used by Bartsch and Wellman (1989): Two tasks required the explanation of a characters action in terms of a false-belief, and two tasks required the prediction of a characters action based on a false-belief. These stories were presented interleaved with other tasks in the testing session. For each story, children were presented with a prototypical box and a matching plain box, and shown that the prototypical box was empty, whereas the plain box contained the item in question. A ller story, in which the prototypical box contained the item and the plain box was empty, also was included to prevent children from guessing that the plain box (rather than the prototypical box) would always be full. A puppet (matching the child for sex and race) was then introduced who had not seen the two boxes before. In the explanation stories, the puppet expressed a desire for the item, and walked to the prototypical box. Children were asked Why is [puppet] looking in there? If no belief-based answer was forthcoming, the prompt What does [puppet] think? was given. Children scored 2 points if they spontaneously referred to a mistaken belief in explaining the puppets action and passed the reality control question (Are there [e.g., crayons] in there really?), 1 point if they referred to a mistaken belief after prompting and passed the reality control question, and 0 points if they made no clear reference to a mistaken belief. In the prediction stories, the puppet expressed a desire for the item, and then walked to a position equidistant between the two boxes. Children were asked to predict where the puppet would look. This question was presented in a forced-choice format; children were coded as successful if they responded correctly to both the test question and the reality control question (Will [puppet] nd the [e.g., cereal] in there?). 2. Unexpected-identity task. Children were told a story from a peep-through pop-up book that involved an element with a deceptive identity, following the procedure used by Hughes, Dunn, and White (1998). On each of the rst ve pages, the child saw an eye peeping through a hole in the page. On the sixth page, children were asked to name what they could see (an eye) before turning over the nal page, which revealed that the eye was actually a spot on a snakes tail. Turning back to the penultimate page,

two test questions were asked. First, children were asked what they had thought the picture was before the page was turned. Second, a naive puppet was introduced, and children were asked to predict whether the puppet would think that the picture was of an eye or a snake. Children were credited with success on the recall-own-false-belief and predict-others-falsebelief questions only if they also passed the reality control questions (What is it really?). 3. Current false-belief tasks. Children heard two stories based on the story developed by Harris et al. (1989). These are designed to examine childrens understanding of a characters false-belief and also how the character will feel as a result of that false-belief. For the purposes of this analysis, however, only the false-belief aspect of the stories was used. The two stories used in the present study involved a character surprising a friend, and placed the false-belief scenario in a realistic context that children would be expected to be familiar with. One story involved a mean trick, in which a character exchanged his friends favorite drink for a disliked drink while the friend was away. The second story involved a nice surprise, in which a character exchanged a friends disliked food for liked food, again in the friends absence. Children were credited with success on each task if they correctly stated what the friend thought he was getting to drink/eat, and what he was really getting. Emotion Understanding Childrens understanding of emotion was tested using the procedure developed by Denham (1986), which involves both affective labeling and affective perspective-taking. Children were shown four felt faces portraying happy, sad, angry, and frightened expressions. They were asked to identify these expressions, rst expressively, by naming, and then receptively, by pointing to the expression the experimenter named. Two points were awarded for correct naming/ pointing, and 1 point for identifying the wrong specic emotion but the correct valence (e.g., calling the frightened face sad or bad), and 0 points for a wrong-valence emotion (e.g., picking happy instead of angry). Childrens scores for expressive and receptive labeling were then summed to give an aggregate score (maximum score 16, from eight identications). To examine affective perspective-taking ability (childrens understanding of the links between particular situations and emotions), children were introduced to a puppet (matching the child for gender and race) with no face. The puppet enacted 16 vignettes portraying situations in which the protagonist felt

Cutting and Dunn

857

happiness, sadness, anger, or fear, such as going to the zoo, seeing a parent off on a trip, having a toy hidden by a sibling, or having a bad dream. Each of the stories was acted out with full vocal and facial cues for the puppets feelings. In eight of the vignettes, the puppets emotion was what most people would be expected to feel in the situation. In the other eight, the puppets emotion differed from the emotion previously predicted by the mother for the child. At the end of each story children were asked how the puppet felt. Children could respond verbally, or nonverbally, by selecting one of four felt faces (children were credited with their best answers if they gave both a verbal and a nonverbal answer and these were contradictorya rare occurrence). Responses were scored as in the affective labeling task (maximum score 32). Language Measures Childrens verbal skills were assessed using measures of both receptive vocabulary (long form of the BPVS; Dunn, Dunn, Whetton, & Pintilie, 1982) and narrative expressive language ability (the Bus Story; Renfrew, 1991). The Bus Story involves the child retelling a simple story about a bus, from a picture book, which is transcribed verbatim and scored for content/coherence, sentence length, and basic grammatical complexity. Family Background Family background data were collected concerning family structure (single or two-parent family), the number of adults in the home, the number of languages spoken in the home, the number of other children in the home, and parental education and occupation. Other children living in the home were almost all siblings (full, half, or step siblings), but in three cases included cousins or foster children living permanently with the family. For simplicity, other children in the home will be referred to as siblings. Twenty-eight of the families in the study (22%) were single-parent families (mothers only). Twenty-two of the families had only one adult living in the home, 13 had three or four adults in the home, and the remainder had two adults at home. Ten families spoke two or more languages at home, but all children were uent in English. Twenty-six of the children in the study were only children or had no siblings living in the home. Sixty-ve children had one sibling living at home, 27 had two, 7 had three, 2 had four, and 1 child had seven siblings at home. Parental educational level was classied according

to the following scheme: 4 degree or higher, 3 up to A level (equivalent to high school graduation in the United States), 2 up to GCSE/O level (national examinations taken at 15/16 years), 1 no qualications. Occupational class was classied using the Standard Occupational Classication (Ofce of Population Censuses and Surveys, 1991), as follows: I professional (e.g., doctor, solicitor), II managerial and technical (e.g., teacher, computer analyst), IIIN skilled, nonmanual (e.g., secretary, nancial clerk), IIIM skilled, manual (e.g., hairdresser, carpenter), IV semiskilled (e.g., gardener, security guard), V unskilled (e.g., cleaner, laborer). In cases where a parent was not currently working, most recent occupation was categorized. Table 1 summarizes the educational and occupational class description of the sample. Missing data include data that could not be collected and absent parents (fathers only). The distribution of occupational classes shown in Table 1 is slightly higher than is reported for the British population (1981 census), but this is largely because parents who were not currently working were classied according to their most recent occupation (these parents would not be included in the occupational census). In fact, 18% of fathers for whom there were data were unemployed. Many mothers were not working either (34%), but this was generally by choice. Of the mothers who were working, 51% worked part-time. Twenty-one percent of families did not have a parent in full-time work. Classifying families more broadly, according to fathers occupational class (mothers in single-parent families), the sample overall was about equally divided between workingclass families (48%; working-class occupational classes IIIV) and middle-class families (52%; middleclass occupational classes I, II). Although it is true
Table 1 Summary of Educational Level and Occupational Class (Percentage of Sample in Each Category) Mother Educational level Degree or higher Up to A level Up to GCSE/O level No qualifications Missing data Occupational class Professional (I) Managerial/technical (II) Skilled, nonmanual (IIIN) Skilled, manual (IIIM) Semiskilled (IV) Unskilled (V) Missing data Father

29.7 20.3 28.9 15.6 5.5 10.2 32.0 32.8 6.3 12.5 1.6 4.7

32.8 14.8 14.1 13.3 25.0 18.8 33.6 7.0 16.4 4.7 4.7 14.8

858
Table 2

Child Development
Intertask Correlations for False-Belief Tasks N % Pass 1 2 3 4 5 6 7 8

Explain action 1. Crayons 2. Plasters Predict action 3. Lego 4. Cereal Unexpected identity 5. Others false-belief 6. Recall own false-belief State others belief 7. Mean trick 8. Nice surprise *p .05; ** p .01.

116 114 117 117 123 123 112 124

56 57 42 36 39 55 50 45

.82**

.31** .40**

.27** .38** .46**

.07 .12 .25** .21*

.05 .11 .16 .01 .18*

.43** .52** .27** .31** .25** .24**

.28** .37** .38** .35** .29** .28** .47**

that social class does not necessarily correlate with income level, in British samples like this, occupational class has proved robust as a marker variable for socioeconomic position (Ofce of Population Censuses and Surveys, 1991, Vol. 3, p. 11). Sibling Relationship Interview For those children with at least one sibling between 2 and 12 years of age (n 92), we interviewed mothers about the quality of the target childs relationship with their nearest sibling, using an interview developed in the Colorado Adoption Project (Stocker, Dunn, & Plomin, 1989). This interview asks about both positive and negative aspects of the sibling relationship and produces two factors (positivity and negativity) for both the target child and the sibling. The positivity factor includes items such as affection and playing together; the negativity factor includes items such as quarrelling and jealousy.

question, but were treated as missing for that task. The intertask correlations for the false-belief tasks are shown in Table 2. Virtually every task correlated signicantly, p .01, with every other task, with the notable exception of recall-own-false-belief in the unexpected-identity task. An irregularity in the presentation of this section of the task was later identied, which seems to account for the lack of correlation between the recall-own-false-belief task and the other false-belief tasks. This task was therefore excluded from further analysis. Cronbachs for the remaining seven false-belief tasks was .80. Each child was given an aggregate false-belief score, calculated as the number of points scored divided by the maximum number of points that could have been scored (excluding missing tasks and tasks in which the child failed the control). A wide range of individual differences was apparent; the mean aggregate false-belief score was .42, with a standard deviation of .32. Emotion Understanding Childrens understanding of emotion was assessed using Denhams (1986) affective labeling and affective perspective-taking tasks. One hundred and four children completed the affective labeling task. The
Table 3 Task BPVS Information Sentence length Complexity Language Scores N 127 122 122 122 Mean 97.4 17.4 8.3 1.0 SD 12.4 8.4 2.4 1.2 Age Norm 100 16 8 1

RESULTS The results are presented in two sections. First, descriptive results for the various measures used are presented. Second, the relations between these measures are explored. False-Belief Tasks Children scored 1 point for each false-belief task passed, except for the two explanation tasks, in which two points were given for a spontaneous correct explanation, and 1 point for a prompted correct explanation. Children who failed control questions (which was very rare) were not counted as failing the task in

Cutting and Dunn

859

of the childs story), (2) sentence length (the mean length of the childs longest ve sentences), and (3) complexity (a measure of the grammatical complexity of the childs sentences in terms of subordinate clauses). The information score is the primary index of narrative expressive language ability. All four language scores correlated signicantly with each other, all ps .01. Correlations between False-Belief Understanding, Emotion Understanding, and Language Childrens false-belief understanding, emotion understanding, and language abilities were clearly related to one another; all three domains were highly correlated. Age also correlated with each domain, but all correlations remained signicant after age was partialed out (see Table 4). Language was highly correlated with both false-belief and emotion understanding, but even with all four language measures and age partialed out, the correlation between falsebelief understanding and affective perspective-taking was signicant, r .22, p .02. Correlations between Family Background and Social Cognition Table 5 shows the correlations between the family background variables and childrens false-belief, emotion understanding, and language scores (only BPVS and Bus Story information scores are included here, as these are the primary indexes of language ability). Note that occupational class is reverse coded so that class I (professional) scores highest, and class V (unskilled) scores lowest. Maternal education and occu-

Note: FB false-belief aggregate, Aff-PT taking, Aff-lab affective labeling. * p .05; ** p .01.

affective perspective-

mean affective labeling score was 14.1 (SD 2.2, range 3 16); Cronbachs .71. The affective perspective-taking task was completed by 115 children. The mean affective perspective-taking score was 27.0 (SD 4.3, range 1032); Cronbachs 0.73. Denham combined the affective labeling and affective perspective-taking scores to give an overall affective aggregate score. Cronbachs for the overall score in this study was .81, but as only 96 of the children received both tasks, it was decided to keep the two scores separate to avoid loss of data. Language Measures Childrens mean scores on the two language measures (BPVS long form, and the Bus Story) are shown in Table 3, along with the standardized norms for children of this age. The BPVS score is a measure of receptive vocabulary. The Bus Story provides three scores reecting narrative expressive language ability: (1) information (a measure of the content and coherence

Table 5 Correlations between Family Background Variables, False-Belief, Emotion Understanding, and Language FB Mother variables Education Occupational class Father variables Education Occupational class Other family variables Family structure Languages spoken Adults at home Siblings at home Aff-PT Aff-lab BPVS Information

.31** .37** .15 .26** .11 .06 .11 .02

.41** .30** .20 .30** .22** .12 .12 .07

.38** .26** .31** .34** .09 .07 .08 .13

.40** .23** .38** .45** .22** .08 .04 .26**

.29** .21* .23* .29** .19* .21** .10 .12 affec-

Note: FB false-belief aggregate, Aff-PT tive labeling. * p .05; ** p .01.

affective perspective-taking, Aff-lab

860

Child Development

pational class were highly correlated with childrens false-belief, emotion understanding, and language abilities. The same was true for paternal occupational class, but not for paternal education (this may simply reect the fact that education data were missing for a quarter of the childrens fathers). The relation between parental occupational class and childrens social cognition was especially clear when group means were examined. Children from middle-class families did signicantly better than children from working-class families in all the domains that we assessed. This is shown in Table 6. In this table, families are classied according to the highest occupational class assigned to the mother or father (middle-class occupational classes I, II; workingclass occupational classes IIIV). This is thus the most generous classication; if families are classed according to mothers or fathers occupational class separately or are given an average occupational class, the difference between working-class and middle-class children is even more pronounced. All the differences shown in Table 6 between middle-class and workingclass children are signicant at p .01 (t tests). This is not simply a function of age (the two groups of children do not differ in age distribution) or language the differences between middle-class and workingclass children for false-belief understanding and affective perspective-taking remain signicant, p .05, even when language is covaried out. Having a single parent was not related to childrens false-belief ability, but children with single parents did perform signicantly worse on the affective perspective-taking and language measures. Bilingual children performed as well as children who spoke only English on all the tasks except the Bus Story, but only 10 children in the sample were bilingual, so this nding should be viewed with caution. The number of adults living in the home was unrelated to childrens social cognition and language skills. The number of siblings children had did not correlate with their false-belief or emotion understanding scores either, a nding that is at odds with some previous research ndings. We examined the effect of younger and older siblings separately, but still found
Table 6 Class Middle (n 76)a Working (n 48)a

no signicant correlations with false-belief or emotion understanding. For those children with at least one sibling between 2 and 12 years of age, we also examined whether the mothers report of the quality of the childs relationship with their nearest sibling was associated with the childs social cognition. The interview used yielded positivity and negativity factors for both the target child and the sibling. None of these factors correlated signicantly with the childrens false-belief and emotion understanding scores, however, although correlations between positivity in the sibling relationship and childrens false-belief scores did increase for children with more siblings. Multivariate Analyses In order to explore further the links between falsebelief understanding, emotion understanding, language, and family background, we conducted a series of linear hierarchical multiple regressions. Initially, on the basis of the correlational results, we entered age, family background (maternal education and occupational class, and paternal occupational class), language (BPVS and Bus Story information), and emotion understanding (affective labeling and affective perspective-taking) into a multiple regression as one step, with false-belief score as the dependent variable. This established that variance in age, family background, language ability, and emotion understanding together accounted for 48.8% of the variance in childrens false-belief scores, R2 .488, F(8, 78) 8.33, p .001. A second regression was conducted for affective perspective-taking, including false-belief score as an independent variable and excluding affective labeling. Affective perspective-taking was used as the dependent variable rather than affective labeling, because it is the more stringent measure of emotion understanding. Variance in age, family background, language ability, and false-belief understanding together accounted for 34.5% of the variance in childrens affective perspective-taking scores, R2 .345, F(7, 96) 6.70, p .001. In further regressions, each variable (or group of variables) was systematically separated out in a second step to determine the unique variance

Mean Scores of Middle-Class and Working-Class Children FB .50 .31 Aff-PT 28.1 25.3 Aff-lab 14.6 13.4 BPVS 101 91 Information 19 15

Note: FB false-belief aggregate, Aff-PT affective perspective-taking, Aff-lab affective labeling. a ns using most generous classication scheme (class data missing for four families).

Cutting and Dunn

861

False-Belief Independent Variables Age Family background Language Emotion understanding False belief R2

.041 5.62* .123 5.59** .120 8.18** .002 .17 Total R2 .488 F(8, 78) 8.33, p .01

.029 3.89* .041 1.87 .072 4.87** .014 1.91 Total R2 .345 F(7, 96) 6.70, p .01

*p

.05; ** p

0.01.

that it contributed to false-belief understanding and to emotion understanding (affective perspectivetaking). Table 7 summarizes the results of both sets of regressions. Variance in language ability contributed the most unique variance to both false-belief and emotion understanding. Age also was a signicant contributor to both areas. Variance in family background contributed uniquely to false-belief ability, but its contribution to emotion understanding was almost all overlapping with variance in the other factors examined. False-belief and emotion understanding did not contribute to each other independently of the other factors included in the regressions. Gender Differences There were some gender differences in our results: girls performed somewhat better than boys on the false-belief tasks (M [girls] .48, SD .34; M [boys] .36, SD .30), and the difference was signicant, t(125) 2.22, p .05. Girls also performed slightly better than boys on both emotion understanding tasks, but neither difference was signicant. There were no signicant differences between girls and boys in any language score, in age, or for any family background variable. The intertask correlations for girls and boys also were very similar, with one interesting exception: the relation between language and false-belief understanding. For boys, all four language measures correlated highly with false-belief scores (r .40, p .01 in each case). For girls, however, the picture was different. The Bus Story information and complexity scores correlated signicantly with false-belief scores (information, r .44, p .01; complexity, r .29, p .05), but neither the BPVS nor the Bus Story sentence length scores did. In fact, the correlations between false-belief and BPVS scores differed signicantly be-

tween boys and girls (boys girls; z 2.04, p .05). Girls BPVS scores did show more variance than did boys, but even when two groups with equal variance in BPVS scores were selected, false-belief understanding did not correlate signicantly with the BPVS for girls, but did so for boys, and these correlations still differed signicantly. This suggests that for boys, all aspects of language may be related to social cognition. For girls, however, it may be that narrative expressive abilities are particularly important. DISCUSSION One of the most important aspects of this study is the nature of the sample itself. The majority of investigations into young childrens social cognition have assessed children from middle-class families, with relatively wealthy and well-educated backgrounds. In this study, the children came from a wide range of backgrounds (see Table 1). A large proportion of the families came from extremely deprived areas of South London, lived in poor housing, and had low incomes. In addition to the broad spectrum of family backgrounds, this study assessed a relatively large number of children (128) and made comprehensive assessments of social cognition, using a multimeasure approach designed to give an accurate picture of each childs ability. The results of the study were compelling. First, we found that family background has signicant implications for young childrens social cognition. Second, our results support previous research demonstrating an important role for language in theory of mind development (Happ, 1995; Jenkins & Astington, 1996). Third, the ndings in some respects support the suggestion that social cognition is not a unitary concept, and that, in particular, understanding of false-belief and understanding of emotion should be viewed as related but distinct aspects of social cognition ( Dunn,

862

Child Development

1995; Dunn et al., 1991). We will discuss each of these points in turn. Family Background and Social Cognition The results of this study show that childrens family background makes a strong contribution to the development of their social cognition, in particular, to their understanding of false-belief. This is in addition to, and independent of, the contributions of language and age. We used very general measures of family background, however, and the task now is to identify exactly how family background affects young childrens understanding of themselves and other people. What are the mechanisms by which individual differences in factors as basic as parental education and occupational class affect young childrens understanding of thoughts and feelings? Is it via the ways in which parents talk to their children and interact with them? Or via the activities they take part in, or facilities they are able to provide for their children? Do different styles of parenting (e.g., authoritarian versus authoritative) inuence young childrens social cognition? Many of these factors are known to vary across different family backgrounds (for a review see Hoff-Ginsberg & Tardif, 1995), but their potential links to childrens social cognition remain largely unexplored (but see Vinden, 1997, for an exception). Of course, none of these possibilities is mutually exclusive, and genetic as well as environmental links may underpin relations between parent and child variables (Scarr, 1992). We found no correlation between childrens falsebelief understanding and the number of siblings (older, younger, or both) that they had. This may in part reect the diverse nature of the sample involved in this studyit is possible that siblings from different family backgrounds interact in qualitatively different ways (Dunn, Slomkowski, & Beardsall, 1994), which may affect theory of mind development. It also should be noted that this lack of correlation between childrens false-belief understanding and number of siblings does not imply that siblings do not inuence young childrens social cognition; indeed, there is plenty of evidence to the contrary (e.g., Azmitia & Hesser, 1993; Brown & Dunn, 1992). We suggest that it is the quality or type of sibling interactions and relationships that are likely to be important for theory of mind developmentnot simply the number of siblings that a child has. As a rst step toward investigating this possibility, we interviewed mothers about the quality of the target childs relationship with their nearest sibling, but this did not correlate with the childrens false-belief scores. We asked only about one

sibling relationship, however; in larger families all the sibling relationships may be important. It also is possible that it is more specic aspects of sibling interactions, rather than global factors, that matter most in the development of social cognition. Language and Social Cognition Several recent studies have highlighted the importance of childrens language abilities in theory of mind development (e.g., Happ 1995; Jenkins & Astington, 1996). Our results strongly support this nding, because language made a signicant contribution to false-belief understanding, independent of age and family background. Our results also extend this nding to emotion understanding. As Jenkins and Astington (1996) point out, we need to know which aspects of language are most strongly related to the development of social cognition. This study provides some pointers. Receptive vocabulary (BPVS) and several aspects of expressive language (Bus Story) all related to the social cognition measures. As such, it may be misleading to look for one specic aspect of language that affects social cognition more than any other. One potentially interesting nding was the difference between girls and boys with regard to language and false-belief understanding. Although all aspects of language related to boys false-belief understanding, narrative ability seemed to be particularly important for girls. This nding requires replication before it is interpreted further, but we speculate that the involvement of language in the development of social cognition may differ for girls and boys. False-Belief and Emotion Understanding Childrens false-belief performance was highly correlated with their understanding of the emotional consequences of events in this study. Correlation does not imply cause, however, and the regression analyses indicated that these two domains of social cognition may in fact be somewhat distinct in children of this age. Language proved to be a strong contributor to both domains, as did age. Family background was clearly related to both false-belief and emotion understanding, but only contributed uniquely to falsebelief understanding. More important, variance in false-belief understanding did not contribute independently to variance in emotion understanding or vice versa. There are several ways in which these ndings can be interpreted. False-belief understanding and emotion understanding may follow different develop-

Cutting and Dunn

863

mental courses; family background, as we measured it, may be important for some aspects of childrens emotion understanding, but at an earlier, or later, phase than we assessed. Alternatively, family background may be important for both domains of social cognition, but its inuence on emotion understanding may be almost entirely overlapping with the inuence of other factors such as language. A third possibility is that childrens family background may be important for both false-belief and emotion understanding, but in different ways; whereas factors such as parental education and occupational class affect childrens understanding of false-belief, more proximal processes, such as family interactions and discourse about feelings, may be important for childrens understanding of the causes of particular emotional experiences (Brown & Dunn, 1996; Dunn et al., 1991). Whichever interpretation is most accurate, our results add weight to the suggestion that false-belief and emotion understanding are distinct aspects of social cognition, especially when taken in conjunction with ndings suggesting that the two domains have differential precursors and sequelae (Dunn, 1995; Dunn et al., 1991). In summary, the results reported in this article emphasize the importance of individual differences when considering the development of young childrens social cognition. In particular, they raise the issue of individual differences in family background and the importance of considering these differences when examining childrens theory of mind. A second issue that arises from our results is the nature of the relation that exists between childrens understanding of false-belief and their understanding of emotion. Our ndings suggest that these domains are to some extent distinct from each other, at least at the age of 4, and that the correlation that exists between them is a result of the inuence of other factors such as age, language, and family background. The picture is still far from clear, however, and there is a need for further investigation into different aspects of young childrens social cognition; we suggest that longitudinal studies of individual differences may be especially helpful. Finally, a word of caution. Generalizations about young childrens social cognition often are made based on samples of children with relatively homogenous backgrounds; our results indicate that clear differences exist between children from different backgrounds, and we would therefore caution against such generalizations. It is common now to partial out both age and language ability when investigating theory of mind ability at the very least, perhaps a measure of family background should be included in this list.

ACKNOWLEDGMENTS This study was supported by a grant to Judy Dunn from the National Institute of Child Health and Human Development (grant 5R01HD23158). We would like to thank all the families and schools who participated in the study, and the researchers and staff involved in the Friendship Project. We would also like to thank Francesca Happ and Claire Hughes for helpful comments on an earlier draft of this article.

ADDRESSES AND AFFILIATIONS Corresponding author: Alexandra L. Cutting, Institute of Psychiatry, Kings College London, Social, Genetic, and Developmental Psychiatry Research Centre, 111 Denmark Hill, London SE5 8AF, UK; e-mail: [email protected]. ac.uk. Judy Dunn is also at the Institute of Psychiatry.

REFERENCES
Astington, J. W. (1993). The childs discovery of the mind. Cambridge, MA: Harvard University Press. Azmitia, M., & Hesser, J. (1993). Why siblings are important agents of cognitive development: A comparison of siblings and peers. Child Development, 64, 430444. Baron-Cohen, S. (1988). Social and pragmatic decits in autism: Cognitive or affective? Journal of Autism and Developmental Disorders, 18, 379401. Baron-Cohen, S. (1991). Do people with autism understand what causes emotion? Child Development, 62, 385 395. Bartsch, K., & Estes, D. (1996). Individual differences in childrens developing theory of mind and implications for metacognition. Learning and Individual Differences, 8(4), 281304. Bartsch, K., & Wellman, H. (1989). Young childrens attribution of action to beliefs and desires. Child Development, 60, 946964. Brown, J. R., & Dunn, J. (1992). Talk with your mother or your sibling? Developmental changes in early family conversations about feelings. Child Development, 63, 336349. Brown, J. R., & Dunn, J. (1996). Continuities in emotion understanding from 3 to 6 years. Child Development, 67, 789802. Cummings, E. M., Hollenbeck, B., Ianotti, R., RadkeYarrow, M., & Zahn-Waxler, C. (1986). Early organization of altruism and aggression: Developmental patterns and individual differences. In C. Zahn-Waxler, E. M. Cummings, & R. Ianotti (Eds.), Altruism and aggression (pp. 165 188). Cambridge, UK: Cambridge University Press. Denham, S. A. (1986). Social cognition, prosocial behavior, and emotion in preschoolers: Contextual validation. Child Development, 57, 194201. de Villiers, J. G. (1995, April). Steps in the mastery of sentence complements. Paper presented at the biennial meeting of

864

Child Development

the Society for Research in Child Development, Indianapolis, IN. Dunn, J. (1995). Children as psychologists: The later correlates of individual differences in understanding of emotions and other minds. Cognition and Emotion, 9, 187201. Dunn, J., Bretherton, I., & Munn, P. (1987). Conversations about feeling states between mothers and their young children. Developmental Psychology, 23(1), 132139. Dunn, J., & Brown. J. R. (1994). Affect expression in the family, childrens understanding of emotions, and their interactions with others. Merrill-Palmer Quarterly, 40(1), 120 137. Dunn, J., Brown, J., Slomkowski, C., Tesla, C., & Youngblade L. (1991). Young childrens understanding of other peoples feelings and beliefs: Individual differences and their antecedents. Child Development, 62, 1352 1366. Dunn, J., Slomkowski, C., & Beardsall, L. (1994). Sibling relationships from the preschool period through middle childhood and early adolescence. Developmental Psychology, 30, 315324. Dunn, L. M., Dunn, L. M., Whetton, C., & Pintilie, D. (1982). British Picture Vocabulary Scale. Windsor, UK: NFER-Nelson. Eisenmajer, R., & Prior, M. (1991). Cognitive linguistic correlates of theory of mind ability in autistic children. British Journal of Developmental Psychology, 9, 351364. Feldman, C. F. (1988). Early forms of thought about thoughts: Some simple linguistic expressions of mental state. In J. W. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind (pp. 126137). New York: Cambridge University Press. Freeman, N. H., Lewis, C., & Doherty, M. J. (1991). Preschoolers grasp of a desire for knowledge in false-belief prediction: Practical intelligence and verbal report. British Journal of Developmental Psychology, 9, 139157. Garner, P. W., Jones, D. C., Gaddy, G., & Rennie, K. M. (1997). Low-income mothers conversations about emotions and their childrens emotional competence. Social Development, 6, 3752. Hadwin, J., & Perner, J. (1991). Pleased and surprised: Childrens cognitive theory of emotion. British Journal of Developmental Psychology, 9, 215234. Happ, F. G. E. (1995). The role of age and verbal ability in the theory of mind task performance of subjects with autism. Child Development, 66, 843855. Harris, P. L. (1994). The childs understanding of emotion: Developmental change and the family environment. Journal of Child Psychology and Psychiatry, 35(1), 328. Harris, P. L., Johnson, C., Hutton, D., Andrews, G., & Cooke, T. (1989). Young childrens theory of mind and emotion. Cognition and Emotion, 3, 379400. Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children. Baltimore: Brookes. Hoff-Ginsberg, E., & Tardif, T. (1995). Socioeconomic status and parenting. In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 2. Biology and ecology of parenting. Mahwah, NJ: Erlbaum.

Holmes, H. A., Black, C., & Miller, S. A. (1996). A cross-task comparison of false belief understanding in a Head-Start population. Journal of Experimental Child Psychology, 63(2) 263285. Hresko, W. P., Reid, D. K., & Hammill, D. D. (1981). The Test of Early Language Development (TELD). Austin, TX: Pro-Ed. Hughes, C., & Dunn, J. (1998). Understanding mind and emotion: Longitudinal associations with mental-state talk between young friends. Developmental Psychology, 34(5), 10261037. Hughes, C., Dunn, J., & White, A. (1988). Trick or Treat? Uneven understanding of mind and emotion and executive dysfunction in hard-to-manage preschoolers. Journal of Child Psychology and Psychiatry, 39, 981994. Jenkins, J. M., & Astington, J. W. (1996). Cognitive factors and family structure associated with theory of mind development in young children. Developmental Psychology, 32(1), 7078. Lewis, C., Freeman, N. H., Hagestadt, C., & Douglas, H. (1994). Narrative access and production in preschoolers false belief reasoning. Cognitive Development, 9(4), 397 424. Lewis, C., Freeman, N. H., Kyriakidou, C., Maridaki-Kassotaki, K., & Berridge, D. M. (1996). Social inuences on false-belief access: Specic sibling inuences or general apprenticeship? Child Development, 67, 29302947. Lewis, C., & Osborne, A. (1990). Three-year-olds problems with false belief: Conceptual decit or linguistic artifact? Child Development, 61, 15141519. Mitchell, P. (1997). Acquiring a conception of mind: A review of psychological research and theory. Hove, UK: Psychology Press. Moore, C., Pure, K., & Furrow, D. (1990). Childrens understanding of the modal expression of speaker certainty and uncertainty and its relation to the development of a representational theory of mind. Child Development, 61, 722730. Moses, L. J., & Flavell, J. H. (1990). Inferring false beliefs from actions and reactions. Child Development, 61, 929945. Murphy, L. B. (1937). Social behavior and child personality: A exploratory study of some roots of sympathy. New York: Columbia University Press. Ofce of Population Censuses and Surveys. (1991). Standard Occupational Classication. London: HMSO. Olson, D. R. (1988). On the origins of belief and other intentional states in children. In J. W. Astington, P. L. Harris, & D. R. Olson (Eds.), Developing theories of mind (pp. 414 426). New York: Cambridge University Press. Perner, J., Ruffman, T., & Leekam, S. R. (1994). Theory of mind is contagious: You catch it from your sibs. Child Development, 65, 12281238. Renfrew, C. E. (1991). The Bus Story: A test of narrative speech. Oxford, UK: Winslow Press. Roazzi, A., & Bryant, P. (1997). Explicitness and Conservation: Social class differences. International Journal of Behavioral Development, 21(1), 5270. Saarni, C., & Harris, P. L. (Eds.). (1991). Childrens understanding of emotion. Cambridge, UK: Cambridge University Press.

Cutting and Dunn

865

Scarr, S. (1992). Developmental theories for the 1990s: Development and individual differences. Child Development, 63, 119. Stocker, C., Dunn, J., & Plomin, R. (1989). Sibling relationships: Links with child temperament, maternal behavior, and family structure. Child Development, 60, 715 727. Sullivan, K., & Winner, E. (1993). Three-year-olds understanding of false mental states: The inuence of trickery. Journal of Experimental Child Psychology, 56, 135148. Vinden, P. G. (1997, April). Parenting and theory of mind. Paper presented at the biennial meeting of the Society for Research in Child Development, Washington, DC. Wahi, S., & Johri, R. (1994). Questioning a universal theory

of mind: Mental-real distinctions made by Indian children. Journal of Genetic Psychology, 155, 503510. Welch-Ross, M. K. (1997). Mother-child participation in conversation about the past: Relationships to preschoolers theory of mind. Developmental Psychology, 33(4), 618629. Wellman, H. M., & Banerjee, M. (1991). Mind and emotion: Childrens understanding of the emotional consequences of beliefs and desires. British Journal of Developmental Psychology, 9, 191214. Youngblade, L. M., & Dunn, J. (1995). Individual differences in young childrens pretend play with mother and sibling: Links to relationships and understanding of other peoples feelings and beliefs. Child Development, 66, 1472 1492.