Psychology Developmental 2001, Vol. 37, No. 5, 587-596
Copyright 2001 by the American Psychological Association, Inc. 0012-1649/01/S5.00 DOI: 10.1037//0012-1649.37.5.587
Fairness or Stereotypes? Young Children's Priorities When Evaluating Group Exclusion and Inclusion
Melanie Killen, Kerry Pisacane, Jennie Lee-Kim, and Alicia Ardila-Rey
University of Maryland
Seventy-two children (35 4!/2-year-olds and 37 5'/2-year-olds), nearly evenly divided by gender, from European American (71%), Asian American (17%), and African American (12%) middle-class backgrounds, were individually interviewed about straightforward exclusion and inclusion for two genderstereotypic peer-group contexts: activities (doll and truck play) and role-play (teacher and firefighter). All children evaluated straightforward exclusion based on gender (e.g., girls excluding a boy from doll play) as wrong and used moral reasons. Preliminary inclusion decisions in the activity contexts (choosing a boy or a girl to join the group) were based on stereotypic expectations, particularly for younger children. Given the opportunity to weigh alternative considerations, however, all children gave priority to fairness over stereotypic expectations in both multifaceted inclusion peer-group contexts.
The present study was conducted to examine young children's judgments about inclusion and exclusion from groups and whether young children give priority to fairness or to stereotypic expectations when asked to make decisions about inclusion in genderstereotypic play situations. Although peer rejection is a wellstudied topic (Asher & Coie, 1990), little is known about how children reason about exclusion from groups when the exclusion is based on group membership. Most studies have investigated peer rejection from the perspective of the child who has been rejected and have focused on the behavioral and social-cognitive differences between high-status and low-status children (for reviews see Coie, Dodge, & Kupersmidt, 1990; Dodge & Feldman, 1990). This individual deficit model has led to interventions aimed at altering the behavior and thoughts of the rejected child, without considering nonbehavioral characteristics and group processes. In a review of the research, Hymel, Wagner, and Butler (1990) argued that the predominance of the social skills deficit model has resulted in a rather limited view of the phenomenon of peer rejection. Hymel et al. suggested the need to expand beyond an individual deficit model and consider group processes in order to fully understand peer inclusion and exclusion. According to Mackie, Hamilton, Susskind, and Rosselli (1996, p. 46), gender is one of the most salient categories for parsing the social environment, particularly in early childhood (see also Maccoby, 1988; Ruble & Martin, 1998). Gender stereotypes for young
children can include judgments about sex-appropriate activities (e.g., girls play with dolls), sex-specific characteristics (e.g., Daddies are strong), and sex-related future roles (e.g., nurses are women) (see Kuhn, Nash, & Brucken, 1978). Stereotypes, commonly defined as cognitive structures that contain an individual's perception of knowledge, beliefs, and expectations about social groups (Ruble & Martin, 1998), are clearly apparent in young children's daily lives. Although research on children's awareness of stereotypes about people on the basis of group membership has shown that gender stereotypes are among the first category of stereotypes to emerge in young children's thinking, most research on stereotypic thinking has not examined whether it is used in morally relevant situations, such as in decisions involving inclusion and exclusion. Yet recent research in the social-cognitive area has demonstrated that children and adolescents bring two types of knowledge to bear on judgments about inclusion and exclusion: moral and social-conventional knowledge (Horn, Killen, & Stangor, 1999; Killen, McGlothlin, & Lee-Kim, in press; Killen & Stangor, 2001; Theimer, Killen, & Stangor, 2001). On the one hand, children and adolescents reason about exclusion by using social-conventional knowledge, such as group norms, expectations, and group identity. These considerations are involved in decisions to exclude someone who does not meet the expectations of the group (e.g., excluding someone from a team who is unqualified). On the other hand, children and adolescents reason about exclusion by using moral knowledge, such as fairness and equal opportunity. There are times when it is unfair or wrong to exclude someone (as when group membership, such as gender and race, is used in an arbitrary manner). Only one study that we know of has examined preschool-aged children's evaluations of exclusion based on gender. In this study, Theimer et al. (2001) interviewed young children regarding straightforward exclusion ("Is it all right to exclude a boy from playing with dolls? Or a girl from playing with trucks?") as well as their decisions about inclusion under different conditions of prior experience with the activity. In Theimer et al.'s study, slightly more than half of the children (65%) judged straightforward ex587
Melanie Killen, Kerry Pisacane, Jennie Lee-Kim, and Alicia ArdilaRey, Department of Human Development, University of Maryland. This research was supported in part by a grant from the National Science Foundation (SBR9729739) and by a project grant from the Graduate Research Board, University of Maryland, College Park. We thank Heidi McGlothlin for technical assistance. In addition, we thank the children, parents, and staff at the Center for Young Children and at the Goddard Child Development Center for participating in the study. Correspondence concerning this article should be addressed to Melanie Killen, Department of Human Development, 3304 Benjamin Building, University of Maryland, College Park, Maryland 20742-1131. Electronic mail may be sent to [e-mail].
588
KILLEN, PISACANE, LEE-KIM, AND ARDILA-REY In the case of inclusion and exclusion, it might be expected that children would view the unfairness of being excluded as being more wrong than the violation of stereotypic expectations because of the psychological harm (e.g., feeling sad) that is inflicted on someone who is excluded. Because inclusion and exclusion decisions involve fairness considerations as well as social group expectations, the social-cognitive domain model provides a framework for investigating how children evaluate these types of decisions (Smetana, 1995; Tisak, 1995; Turiel, 1998). One way to do this is to describe probes to children that emphasize moral (e.g., fairness) or social-conventional (e.g., stereotypic expectations) considerations and to evaluate whether these types of probes influence children's inclusion and exclusion decisions and justifications. In the cognitive area, researchers have emphasized the suggestibility of young children and have suggested that young children are highly susceptible to adult probes (e.g., Ackil & Zaragoza, 1995; Winer & McGlone, 1993). We predicted that children would not be susceptible to any probe but rather would be more influenced by probes emphasizing moral considerations than those emphasizing social-conventional expectations. Central to this study, then, was the prediction that children would give priority to fairness more often than to social expectations and stereotypic norms in multifaceted situations involving inclusion and exclusion. Third, Theimer et al.'s (2001) study was conducted with European American participants (94%) in a homogeneous European American town located an hour outside of a large mid-Atlantic city. The present study interviewed children who lived in a middleclass, mixed-ethnicity suburb of a large mid-Atlantic city, and this sample more closely reflects the general U.S. population (71% European American, 17% Asian American, and 12% African American). The study was not designed, however, to make direct comparisons between children living in different types of communities. Instead, the study was designed to assess children's judgments about exclusion with a sample that reflected the general U.S. population. Finally, Theimer et al. (2001) did not examine age-related changes within the preschool period (only one age group--4-yearolds--was tested). Given that stereotypic expectations about play activities change from 3 to 5 years of age and that children's use of moral and social-conventional justifications changes between 3!/2 and 5lA years of age, we predicted that children's inclusion and exclusion decisions would change during this period as well. In their extensive review of the literature, Ruble and Martin (1998) stated that "stereotype knowledge of child and adult activities and occupations increases rapidly between 3 to 5, reaching a high level by kindergarten" (p. 946). For example, Reis and Wright (1982) examined children's knowledge of adult sex-role stereotypes at three ages--3V4, 4Vi, and 5'/2 years--and found that sex stereotyping emerged at 3Vi years of age and increased during the preschool period. Further, there is a dramatic increase in gender-typed preferences (Ruble & Martin, 1998) and sex-segregatedplay (Maccoby, 1988) during the preschool period. Gender-consistent preferences are over 80% at 4 years of age and peak at 100% by 7 years of age (Ruble & Martin, 1998, p. 953). Sex segregation increases significantly from 4 to 6 years of age (Maccoby, 1988). Whereas 4-year-old children interact with same-sex children 3 times more
elusion made on the basis of gender as wrong. When told that two children had the same prior experience with the activity (e.g., "If a girl and a boy both played with dolls before, whom should the group pick to join the play?"), the majority of children chose the child who fit the stereotype, and they used social-conventional reasons for their decisions. However, when told that the child who fit the stereotype had engaged in the activity before, then the majority of children chose the child who did not fit the stereotype to be included in the play activity, and they gave moral reasons (such as turn-taking) for their decision. Thus, in some contexts children relied on stereotypic expectations to make inclusion decisions, and in other contexts they relied on fairness judgments. There were several limitations of Theimer et al.'s (2001) study, however, and the present experiment was designed to extend that study in several ways. First, Theimer et al. did not analyze children's inclusion judgments when no other information was provided. Children were asked whom a group should pick, a girl or a boy, when the boy and girl had (a) equal prior experience with the activity or (b) unequal prior experience. No assessments were made about children's inclusion decisions without other information (e.g., "Whom should the group pick?"). These results could suggest that "all things being equal," children use stereotypes to make inclusion decisions. Yet children were not asked to make inclusion judgments in the absence of information about prior experience. Second, no information was obtained regarding the extent to which children held onto their beliefs when confronted with competing considerations. For example, would children who picked a boy to play with dolls change their judgment when asked to think about stereotypic expectations about doll play? Alternatively, would children who picked a girl to play with dolls change their judgment when asked to think about issues related to equal opportunity and fairness? Assessing children's reactions to competing considerations provides knowledge about young children's priorities--that is, given two competing considerations (fairness and stereotypes) to contemplate, which consideration is viewed as most important or salient? Using the social-cognitive domain model, some studies have provided evidence that children and adolescents are often able to recognize different social domains (e.g., moral, socialconventional) in multifaceted situations and that the domain that is given priority depends on the salience of the situation and the experience of the individual (Smetana, 1983; Turiel, 1983; Turiel & Smetana, 1984). Which domain is given priority in multifaceted situations involving gender stereotypes and exclusion from groups is not known. Because research has demonstrated how salient gender stereotypes are in the lives of young children, it might be expected that young children would focus more on the socialconventional expectations of gender roles in these types of situations (see Carter & Patterson, 1982; Stoddart & Turiel, 1985) even when confronted with moral considerations, such as fairness. However, numerous studies have demonstrated the salience of the moral domain by showing that young children pay more attention to moral transgressions than to social-conventional transgressions in their actual social interactions (Smetana, 1995). Children judge fairness to be generalizable and unalterable, whereas they judge conventional considerations, such as stereotypic expectations, to be more context specific.
FAIRNESS OR STEREOTYPES?
589
often than with opposite-sex children, 6-year-old children interact with same-sex peers 11 times more often than with opposite-sex peers. Thus, children's play changes from being primarily mixedsex in early preschool (3'/2-41/2 years of age) to primarily sex segregated by late preschool (4V2-5V2 years of age). Few studies have investigated how children's increasingly sex-segregated play bears on their social judgments about inclusion in peer-play situations. Does the fact that girls increasingly play dolls only with girls mean that they would increasingly be willing to exclude a boy from playing with dolls? Similarly, are boys increasingly willing to exclude a girl from playing with trucks? How do young children's stereotypic judgments about play behavior influence their decision making that involves inclusion and exclusion during the preschool period? Because young children increasingly engage in sexsegregated play during the preschool years, it follows that they may use these experiences to justify excluding other children from peer-group play. Yet research has demonstrated that children's ability to justify their moral judgments increases during the preschool years (Killen, Breton, Ferguson, & Handler, 1994; Killen & Smetana, 1999; Tisak, 1995). Children around 3'/2 years of age judge moral transgressions to be wrong but give fairly global justifications (e.g., "It's not nice"), whereas by 5 years of age, children give more elaborated justifications (e.g., for hitting: "He would feel hurt just like I feel hurt when someone hits me and so it's wrong to do that to someone else and he would feel very sad"). On the basis of these age differences, we expected that children's justifications for their decisions about inclusion and exclusion would change during the preschool period. In general, the present study was designed to investigate young children's judgments about straightforward exclusion as well as their willingness to change their inclusion judgments when asked to consider competing considerations. We asked 4Vi- and 5lAyear-old children to evaluate four exclusion scenarios: girls excluding a boy from playing with dolls, boys excluding a girl from playing with trucks, girls excluding a boy from role-playing a teacher, and boys excluding a girl from role-playing a firefighter. These scenarios were chosen because previous research has demonstrated that young children view these activities as gender stereotypic (Katz & Ksansnak, 1994; Kuhn et al., 1978). Children were asked to judge a straightforward exclusion scenario and then to select either a girl or a boy to join the group of same-sex children engaged in a play activity consistent with gender stereotypes (it was explained that there was only room for one more child to join the group). On the basis of children's choices, one of two verbal counterprobes was then described. The moral counterprobe, given to children who picked a same-sex child (one who fit the stereotype), emphasized allowing the opposite-sex child a chance to play at the activity (fairness and equal opportunity). The social-conventional counterprobe, given to children who picked the opposite-sex child (who did not fit the stereotype), emphasized the social expectations of same-gender play (stereotypes and social expectations). Children were asked for their choices and justifications before the probe (preprobe choice) and following the counterprobes (postprobe choice). We hypothesized that the majority of children would judge straightforward exclusion based on gender as wrong by using moral reasons (as shown in previous findings by Phinney & Cobb, 1996; Theimer et al., 2001). Prior findings on gender-stereotypic
knowledge suggest that gender stereotypes overwhelm young children's judgments even in straightforward exclusion scenarios. Therefore, it was important to determine whether children viewed straightforward exclusion as wrong from a moral viewpoint. Our expectations regarding children's inclusion judgments (i.e., whom they would pick to join the group) was that younger children would be more likely than older children to pick someone who fit the stereotype. This was based on prior findings demonstrating that gender-segregated play increases with age (Maccoby, 1988; Ruble & Martin, 1998). We also had predictions about children's judgments regarding two play contexts: activities (doll play, truck play) and roleplaying (teacher, firefighter). Prior research has documented that sex-typed toy preferences emerge as early as 3 years of age (Weinraub et al., 1984) and that knowledge about adult sex-role occupations changes during the preschool years (Reis & Wright, 1982), declining by elementary school age (Garrett, Ein, & Tremaine, 1977). In general, Ruble and Martin (1998) reported that gender-stereotypic knowledge about activities emerges prior to gender knowledge about roles and that gender knowledge about roles becomes increasingly flexible with age (although it is not known whether gender knowledge about roles changes during the preschool period). On the basis of these findings, we expected that younger children's stereotypic choices would be more frequent for activities than for roles and that older preschool children would make fewer stereotypic choices in both contexts. Because younger preschoolers' knowledge is less stable, we predicted that they would more likely be influenced by counterprobes for their choices than would older children. In addition, we expected that younger preschool children would use more social-conventional justifications than would older preschool children given that prior research has shown that the use of social-conventional transgressions decreases with age (Killen et al., 1994) and that children evaluate stereotypes using social-conventional justifications (Carter & Patterson, 1982). We also tested for in-group favoritism by gender (whether girls would be more likely to exclude boys than girls or whether boys would be more likely to exclude girls than boys). Some studies have documented in-group biases (Martin, 1989; Powlishta, 1995) and others have found little evidence for it (Killen & Stangor, 2001; Theimer et al., 2001). Because previous studies have shown mixed results, it was an open question whether we would find in-group favoritism in children's evaluations of exclusion by gender. Method
Participants
Participants included 72 children enrolled in two child-care centers in middle-class, mixed-ethnicity suburbs of a large mid-Atlantic city. There were thirty-five 4'/2-year-old children (M - 55.60 months, SD = 3.90, range = 49-61 months), including 16 girls and 19 boys, and thirty-seven 5'/2-year-old children (M = 68.30 months, SD = 5.39, range = 62-81 months), including 17 girls and 20 boys. The children were primarily from middle-class backgrounds, as determined by the school directors, and reflected European American (71%), Asian American (17%), and African American (12%) backgrounds. Parental permission forms were distributed at school, and all children who were given parental permission to participate were included in the study (85%).
590 Procedure
KILLEN, PISACANE, LEE-KIM, AND ARDILA-REY child of the opposite sex. On the basis of the child's preprobe choice, one of two probes was introduced. The moral probe was used if the child initially selected the child who fit the stereotype ("What if some other girls say that they should pick Tom because boys don't usually get a chance to play with dolls and it would be more fair to give Tom a chance to play with the dolls?"). The social-conventional probe was used if the child initially selected the child who did not fit the stereotype ("What if some other girls say that they should pick Sally because boys don't usually play with dolls and doll playing is something that girls do?"). The participant was then asked to reevaluate the choice of whom to include (the postprobe choice: "Should the girls/boys pick Tom/Sally then?") and to justify his or her response. Thus, there were three dependent measures: (a) judgments (all right or not all right to exclude), (b) choice (whom to pick to join the group), and (c) justifications (reasons for judgments and choices).
Children were interviewed individually for approximately 20 min by one of two graduate research assistants trained in interviewing young children. Four exclusion stories were read to all children. Two stories described stereotypic play activities (doll playing and truck playing) and two stories described stereotypic future roles (being a teacher or being a firefighter) (Albert & Porter, 1983; Kuhn et al., 1978; Theimer et al., 2001). Brightly colored Wi X 11 in. (24 X 28 cm) picture cards were used to aid in comprehension. Interviews were recorded on audiotape and transcribed for analysis by the two graduate research assistants. A within-subjects design was used (all children evaluated all four stories). First, there was an evaluation of exclusion when there were no competing considerations, which was referred to as straightforward exclusion. For example, for the doll-playing story, the participant was told a story about a boy who wants to join a group of girls playing with dolls. The participant was asked if it was all right or not all right (Judgment) for the girls who are playing with the dolls to tell the boy that he cannot play with them because he is a boy and to provide a justification for his or her answer. For example: A group of girls is playing with dolls. John comes over and asks if he can play. Two of the girls say that John cannot play because he is a boy. Is it all right or not all right for the girls to tell John that he cannot play? Why or why not? Second, using the same scenarios described in the straightforward exclusion context, we asked children to select either a boy or a girl to join the group. The participant was asked to decide if he or she thought that the group should select the girl or the boy to join the group given the condition that only one more child could play and to justify their response (preprobe choice): Let's say that two children, Tom and Sally, come over and want to play with the girls. There is only one doll left. The group has to decide whom to pick. Whom should the group pick? How come they should pick him/her? Next, contingent upon the child's choice of either the stereotypic or the nonstereotypic child, the participant was asked to consider selecting the
Coding and Scoring Responses
The judgment responses of all right to exclude received a 0, and the responses of not all right to exclude received a 1 (as has been done in prior studies; see Killen & Smetana, 1999, and Nucci & Weber, 1995, for similar coding procedures). The choices made by children were coded as a 0 when the stereotypic child (e.g., the boy in the truck story) was chosen and as a 1 when the nonstereotypic child (e.g., the girl in the truck story) was chosen. Justifications were coded as moral (prosocial, fairness, equality), socialconventional (social expectation, group identity), or other on the basis of previous research (Killen et al., 1994; Smetana, 1995; Theimer et al., 2001). Each category is described in Table 1. Gender stereotypes were coded as social-conventional justifications (Carter & Patterson, 1982; Theimer et al., 2001). Justifications were scored dichotomously, with a score of 1 indicating that the category was used and a score of 0 indicating that the category was not used (for each category for each participant). Analyses of variance (ANOVAs) rather than log-linear modeling were conducted on these proportions because of the use of a repeated measures design. Previous studies have demonstrated that ANOVA-based procedures are robust when used with dichotomous data (Gaito, 1980) and that log-linear modeling is problematic with repeated measures designs, thus supporting the use of ANOVAs with these types of data (see Kahn, 1999;
Table 1 Coding Categories for Justifications
Category Moral Prosocial Equality and fairness Justification Appeals to helping others (e.g., "Because they can teach him how to play"), being nice (e.g., "Because you should be nice to other people"), and showing empathy (e.g., "Because she will be sad if they don't pick her"). Appeals to maintaining equality and fairness in the treatment of others (e.g., "If they don't let her play, it wouldn't be fair; because boys can play with dolls, too"). Includes issues of equal access (e.g., "Because both boys and girls can both play and you should let them all play together"), and equal opportunity (e.g., "Because there are already three girls and no boys, and boys should get a chance to play with dolls, too"). Appeals to the expectations of the group (e.g., "Boys don't play with dolls"; "Only girls can be teachers"; "Because the game is only for boys"). Appeals to the maintenance of the group (e.g., "Boys like to play with boys"; "Girls don't like to play with boys"; "They should pick the girl because they are all girls"). Unelaborated justifications such as "It's bad," "It's okay," or "I don't know" or other justifications that were too low in frequency to constitute an independent category.
Social-conventional Social expectation Group identity Other Undifferentiated and other
FAIRNESS OR STEREOTYPES? Smetana, 1995; Wainryb, Shaw, Laupa, & Smith, 2001). Chi-square statistics were conducted to assess the type-of-change variable, as described below.
591
discuss the results for the choices and justifications prior to and following the verbal probe.
Reliability
Reliability coding was conducted on the justification data and was calculated with 50% of the interviews (102 data points). Interrater agreement in scoring justifications using Cohen's kappa was .86 (agreement = 95%).
Straightforward Exclusion Measure
A 2 (age) X 2 (context: activities, roles) repeated measures ANOVA for judgments revealed that there were no context or age differences in the way children judged exclusion for the straightforward exclusion measure. As predicted, the vast majority of children (M = .87, SD = .26) judged it to be wrong to exclude someone from an activity or from role-play. Analyses of the justifications given by children for their straightforward exclusion judgments showed that the vast majority of the children used moral justifications (M = .84, SD = .33) and that very few used socialconventional reasons (M = .10, SD = .24). Separate 2 (age) X 2 (context) repeated measures ANOVAs conducted on the moral and social-conventional justifications data revealed that there were no age, context, or Age X Context interaction effects.
Story Order
The two activity stories (dolls and trucks) were presented before the two future-roles stories (teacher and firefighter) because previous research has found that children are more likely to invoke stereotypes when judging exclusion from play activities than from future roles (Theimer et al., 2001); thus, the stories that were more likely to invoke stereotypes were described first in order to decrease social desirability responding. Within each context, the story order was counterbalanced. Analyses indicated that there were no story-order effects.
Results Plan for Analysis
All dependent measures were analyzed with repeated measures ANOVA procedures. In all subsequent analyses, we performed post hoc comparisons with the Bonferroni t test to correct for Type I error. ANOVAs were conducted on proportions of judgments, choices, and justifications. Initial 2 (gender) X 2 (age) X 4 (stories) repeated measures ANOVAs conducted on each one of the questions revealed no gender effects; thus gender was not included in further analyses. The same procedure was used to test for ethnicity, and no results were found, thus excluding ethnicity from further analyses. Because these analyses revealed a story effect, follow-up t tests were conducted on all possible combinations of the four original stories for each question. The analyses revealed no differences between the two stories within each context (e.g., between dolls and trucks within the activity context or between teacher and firefighter within the role-playing context), but significant differences were found for stories across contexts. Thus, the data were analyzed by context (activities or roleplaying). We first report the results for judgments and justifications for the straightforward exclusion measure, and then we
Inclusion Choices
As shown in Table 2, less than half of the children initially chose the stereotypic child to be included in the peer play for doll and truck playing (M = .44), and more than half chose the stereotypic child for the teacher and firefighter role-playing (M = .58). A significant shift occurred when children made their postprobe choices, such that the majority of children chose the nonstereotypic child (M = .61) to be included in the peer play. A 2 (age) X 2 (context: activities, roles) X 2 (choice: preprobe, postprobe) ANOVA with repeated measures on the last two factors was conducted and revealed a main effect for context, F(l, 70) = 7.68, p < .01, choice, F(l, 70) = 4.61, p < .05, and a nearly significant effect for age, F(l, 70) = 3.62, p < .06. As hypothesized, more children chose the nonstereotypic child for the roles (M = .60, SD = .40) than for the activities (M = .52, SD = .40), more children chose the nonstereotypic child for their postprobe choice (M = .61, SD = .39) than for their preprobe choice (M = .51, SD = .38), and younger children chose the stereotypic child (M =.52, SD = .30) more often than did older children (M = .62, SD = .29). Because we had hypotheses about age, we report the nearly significant effect. There were significant no interaction effects.
Table 2 Mean Proportions of Preprobe and Postprobe Nonstereotypic Choices by Age Group
Choice inclusion Preprobe choice Age (in years)
4'/2 5'/2
Postprobe choice M .44 (.33) .59 (.29) .51 (.32) Activities .63 (.37) .58 (.36) .60 (.37) Roles .54 (.41) .69 (.38) .62 (.40) M .59 (.27) .64 (.29) .61 (.28)
Activities .36 (.36) .53 (.35) .44 (.36)
Roles .51 (.41) .65 (.37) .58 (.40)
M
Note. Standard deviations are in parentheses. Choice: 0 = stereotypic child; 1 = nonstereotypic child. N = 72 children (thirty-five 4'/2-year-olds and thirty-seven 5'/2-year-olds).
592
KILLEN, PISACANE, LEE-KIM, AND ARDILA-REY
These findings confirmed our hypotheses that stereotypes would be more prominent for choices about whom to pick for an activity than about whom to pick for role-play and that children's choices shifted from the preprobe to the postprobe decision in only one direction. There was an increase in the tendency to pick the child who did not fit the stereotype, and there was not an increase in the tendency to pick the child who fit the stereotype. Furthermore, younger children were more likely to pick a child who fit a stereotype (than who did not) than were older children.
Justifications for Inclusion Choices
Analyses of children's use of justifications paralleled the findings for their choices of whom to picfc (see Table 3). A 2 (age) X 2 (context: activities, roles) X 2 (choice: preprobe, postprobe) X 2 (justifications: moral, social-conventional) ANOVA with repeated measures on the last three factors revealed that there were significant justification effects, F(l, 70) = 7.16, p < .01, as well as Age X Justification, F(l, 70) = 5.10, p < .05, Context X Justification, F(l, 70) = 12.45, p < .01, and Choice X Justification, F(l, 70) = 6.65, p < .05, interactions. The Age X Justification effect indicated that younger children used more social-conventional justifications overall (M = .46, SD = .31) than did older children (M = .29, SD = .28), r(70) = 3.04, p < .003; there were no age differences for the use of moral justifications. More moral justifications were given for the role stories (M = .57, SD = .43) than for the activity stories (M = .45, SD = .38), t(7\) = 3.77, p < .0001, whereas more social-conventional justifications were given for the activity stories (M = .42, SD = .36) than for the role stories (M = .33, SD = .39), r(71) = 2.96, p < .004. The Choice X Justification interaction showed that there were no differences in the use of moral or social-conventional justifications for the preprobe choice of inclusion but that more moral (M = .56, SD = .41) than socialconventional (M = .32, SD = .37) justifications were given for the postprobe choice of inclusion, /(71) = 3.93, p < .0001. Also, moral justifications were used more often for the postprobe choice (M = .56, SD = .41) than for the preprobe choice (M = .47, SD = .39), t(7\) = 2.37, p < .02, whereas social-conventional justifi-
cations were used more often for the preprobe choice (M = .43, SD = .38) than for the postprobe choice for whom to include (M = .32, SD = .37), f(71) = 2.66, p < .01. As predicted, children used more moral justifications for whom to pick for activities for their postprobe choice than for their preprobe choice, and children used fewer social-conventional justifications for their postprobe choice than for their preprobe choice (see Table 3). The use of justifications for roles did not change from the children's preprobe to postprobe choice but, for both choices, children used more moral than social-conventional justifications. These results confirmed our hypotheses that children would initially view choices about whom to pick for activities in social-conventional terms more often than they would view choices about whom to pick for roles in social-conventional terms and that moral justifications would be more prominent after counterprobes than would social-conventional justifications.
Predictions About Change
Four new variables were created to reflect the different patterns of change (called type of change) for children's choices from preprobe to postprobe for both contexts (activities and roles). These type-of-change variables were as follows: (a) stereotypic child chosen at preprobe/stereotypic child chosen at postprobe (SS); (b) nonstereotypic child chosen at preprobe/nonstereotypic child chosen at postprobe (AW); (c) nonstereotypic child chosen at preprobe/stereotypic child chosen at postprobe (NS); and (d) stereotypic child chosen at preprobe/nonstereotypic child chosen at postprobe (SA^. The first two variables (SS and NN) reflected no change in children's choices (children did not change their choice after hearing the counterprobe). The latter two variables (NS and SN) reflected a change in children's choices (children changed their choice after hearing the counterprobe). Chi-square statistics were computed for the type-of-change variable for each story within each context (i.e., activities and roles). The counts, expected frequencies, and percentages are displayed in Table 4. For the activities/dolls story, a significant chi-square indicated that more children changed than expected by chance, l, N = 72) = 7.65, p < .006. As shown in Table 4, more
Table 3 Mean Proportions of Justifications of Preprobe and Postprobe Choices by Age Group
Choice inclusion Preprobe choice Age and justification 4'/2-year-olds M S-C 5 '/2-year-olds M S-C Total M S-C Activities .33 (.36) .60 (.38) .42 (.36) .43 (.36) .38 (.36) .51 (.38) Roles .51 (.41) .46 (.41) .60 (.44) .23 (.35) .56 (.42) .34 (.39) M .42 (.34) .53 (.34) .51 (.33) .33 (.27) .47 (.33) .43 (.32) Activities .53 (.38) .37 (.35) .54 (.40) .27 (.35) .53 (.39) .32 (.35) Postprobe choice Roles .56 (.42) .41 (.41) .62 (.43) .23 (.37) .60 (.42) .32 (.35) M .54 (.27) .39 (.28) .58 (.35) .25 (.28) .56 (.31) .32 (.29)
Note. Standard deviations are in parentheses. Justifications: M = moral; S-C = social-conventional, N = 72 children (thirty-five 4'/2-year-olds and thirty-seven 5'/2-year-olds).
FAIRNESS OR STEREOTYPES?
593
Table 4 Number of Children Using Each Type of Change Category for Two Contexts
Type of change No change Variable SS NN Activities Dolls No. of children Expected value % Trucks* No. of children Expected value % 10 15.5 14 10 15.9 14 13 18.5 18 14 19.9 19 Roles Teacher No. of children Expected value % Firefighter No. of children Expected value % 9 12.0 13 9 10.9 13 22 25.0 30 25 26.9 35 18 15.0 25 19 17.1 26 23 23.0 32 19 17.1 26 72 16 10.5 22 21 15.1 29 33 27.5 46 27 21.1 38 72* NS Change SN Total
judge it "more all right" to exclude boys than girls; boys did not judge it "more all right" to exclude girls than to exclude boys. Discussion Though previous research has emphasized the salience of stereotypes in the lives of young children (Carter & Patterson, 1982; Ruble & Martin, 1998), the results of this study demonstrated that young children judged straightforward exclusion to be wrong and used moral justifications, even in play contexts in which genderstereotypic expectations are high and gender-segregated play is the norm (Maccoby, 1988). Research has demonstrated that adolescents also reject straightforward exclusion on the basis of group membership (by gender and race, see Killen & Stangor, 2001; by ethnicity, see Phinney & Cobb, 1996), but only one study known to us has shown that young children judge straightforward exclusion by gender to be wrong (see Theimer et al., 2001). We replicated Theimer et al.'s finding with a sample that more closely reflected the general U.S. population. Our central hypothesis was that the moral dimension of multifaceted decisions involving inclusion and exclusion would be more salient to children than the social-conventional dimension. This hypothesis was confirmed. Moral probes focusing on fairness had a greater influence on children's decisions regarding whom to include than were social--conventional probes focusing on genderstereotypic expectations. No study known to us has reported this type of finding. Young children who initially focused on the social-conventional aspects of inclusion changed their decision to focus on moral concerns when they were provided with an opportunity to weigh both stereotypic and moral considerations. In contrast, children who initially focused on the moral dimensions were less likely to change their judgment even when stereotypic expectations were made salient. Though young children are not always able to consider the competing demands of a situation simultaneously, the introduction of alternative perspectives may enable children to weigh multiple considerations and give priority to morality. Although it might be expected that children would be influenced by counterprobes regardless of the point of view being advocated, our findings did not support this. In the activities context (dolls, trucks), few children who picked the stereotypic child refused to change their decision after contemplating the moral considerations. Moreover, for the children who changed their choice in the activities context, more children tended to pick the nonstereotypic child than to pick the child who fit the stereotype. This finding was significant because previous research in the cognitive area (e.g., Ackil & Zaragoza, 1995; Winer & McGlone, 1993) has emphasized young children's suggestibility to adult probes. The results of the present study showed that children did not blindly agree to any countersuggestion. Instead, when children were given the opportunity to consider alternative points of view, moral considerations were more salient than were social-conventional ones. Additional research with more varied probes and diverse scenarios is needed to fully understand young children's susceptibility to probes and countersuggestions. Our age-related findings were that 41/2-year-old children more often chose the stereotypic child prior to hearing the probe than did 5'/2-year-old children and that younger children used more socialconventional justifications than did older children (there were no
72**
72
Note. SS = stereotype to stereotype; NN = nonstereotype to nonstereotype; NS = nonstereotype to stereotype; SN = stereotype to nonstereotype. N = 72 children (thirty-five 4'/2-year-olds and thirty-seven 5Vi-year-olds). * p < .006. ** p < .005.
children changed their choice from stereotypic to nonstereotypic child (n = 33) than from nonstereotypic to stereotypic child (n = 16). Similarly, for the activities/trucks story, more children changed their choice than expected by chance, ^ ( 1 , N = 72) = 7.98, p < .005. As shown in Table 4, more children changed their choice from stereotypic to nonstereotypic child (n = 27) than did not change their choice. (Chi-square tests on these data were not conducted for age effects because there were cells with less than 5 for the expected value.) The chi-square tests for the teacher and firefighter role stories were not significant. As shown in Table 4, for the roles stories, few children chose the stereotypic child and did not change their choice (ns = 9 for both stories). Children either chose the nonstereotypic child and did not change (ns = 22 and 15 for teacher and firefighter, respectively), or they changed their choice as a function of the alternative consideration. These findings indicated that in the activity context, more children changed their score (n = 49) than did not (n = 23), and when they changed they tended to pick the nonstereotypic child (than the stereotypic child). However, for the roles context, most children either initially picked the nonstereotypic child (and did not change) or they changed their decision after hearing a probe. Thus, for the context in which children's play behavior has been shown to be fairly stable with age (sex-segregated activities), the probes were more influential in the expected direction than they were for the context in which studies have documented increased flexibility in the related stereotypes (roles) during elementary school. There were no significant findings regarding in-group favoritism, as evidenced by the lack of gender findings. Girls did not
594
KILLEN, PISACANE, LEE-KIM, AND ARDILA-REY in the preprobe choice context, in which children had to decide between selecting a child who did or did not fit the stereotypic expectations. More than half of the children in the preprobe activity choice context picked the child who fit the stereotype for social-conventional reasons. Justifications focused on stereotypes, expectations, and shared norms (e.g., "Because the dolls are for the girls and not for the boys," and "Because the boys want to play with a boy"). Some children also made inferences about the different capabilities and interests of boys and girls. An example of this, as articulated by one child, was "Because three girls are playing, and boys play the wrong way" or "Boys don't like dolls." These findings emphasize that gender stereotypes influence young children's judgments, particularly their inclusion judgments. Children evaluated the inclusion decisions in the activity context (dolls, trucks) differently from decisions in the role-playing context (teacher, firefighter). Children selected the child who did not fit the stereotype for role-playing such roles as teacher and firefighter more often than for activities such as doll and truck play. These findings provide some preliminary evidence of the relationship between children's stereotypes and judgments about inclusion and exclusion. For scenarios in which it is known that children have strong stereotypes, such as in doll and truck play, children were more likely to select a child who fit the stereotype than a child who did not. However, for scenarios about role-playing, in which gender flexibility has been reported with children of elementary school age (Ruble & Martin, 1998), children were less likely to select a child who fit the stereotype. More research using direct analyses between stereotyped knowledge and inclusion choices is needed to validate this interpretation of the results. Children judged exclusion in the straightforward exclusion context in much the same way that previous studies have documented children's evaluations of prototypical moral transgressions. Typically studies find that between 95% and 100% of children regard unambiguous moral transgression (such as unprovoked hitting or not sharing) as wrong (Smetana, 1995; Tisak, 1995). The present study found that 87% of young children judged exclusion in the straightforward exclusion context as wrong. These findings might be a result of many factors, including exposure to nonstereotypic curriculum, school philosophies, heterogeneity of the communities, and general awareness of stereotypes. These results are somewhat different from the findings of Theimer et al.'s (2001) study in which preschool-aged children were asked about exclusion on the basis of gender, and 65% of the children judged straightforward exclusion as wrong. The Theimer et al. study was conducted in a homogeneous, European American, middle-class community located about an hour outside of a metropolitan region. The present study was conducted in a mixedethnicity, middle-class community within a metropolitan area. It is possible that the heterogeneity of the community of this sample contributed to the greater percentage of children judging straightforward exclusion as wrong. However, there were other differences between the two samples, and without direct analyses of the community variable, this interpretation remains speculative. Reviews of intergroup contact on reducing prejudice based on group membership (such as gender and race), however, have shown that cross-race friendships are significantly related to prejudice reduction and group biases (Pettigrew & Tropp, 2000). Although this literature has more often been focused on race than on gender, the investigators studying this phenomenon predict that intergroup
age differences for the use of moral justifications). The age-related findings for the choice of whom to play with have to be interpreted with caution, however, as this was a nearly significant result (p < .06). Perhaps a stronger age-related pattern of responses might be detected when a larger age range (3'/2-5V2 years of age) is involved. One difficulty with this is that children below 4lA years of age do not consistently provide justifications for their judgments and, in fact, give a high number of undifferentiated reasons (see Killen et al., 1994; Tisak, 1995). This suggests that age differences regarding inclusion choices need to be assessed by means of a multimethod approach--one that relies on indices other than justification responses--for determining age-related patterns regarding orientations to inclusion and exclusion. The age-related findings for justifications support the choice pattern, and together these results indicated that changes are taking place during the preschool years regarding how young children apply their stereotypic knowledge to peer-play decisions. Younger preschool children were more likely to base their inclusion judgments on stereotypic reasons and expectations (e.g., "dolls are only for girls" or "trucks are for boys") than were older preschool children, contrary to what may be expected on the basis of the findings that sex-segregated play increases with age (as reported by Ruble & Martin, 1998). Observational as well as longitudinal research will help fill in the gaps in our knowledge about agerelated patterns during the preschool period. In general, children's justifications also revealed the sophistication by which they conceptualized the moral components of the scenarios presented to them. To explain why exclusion was wrong in the straightforward exclusion scenario, children used a variety of moral reasons, including prosocial reasons, fairness, and equality. For example, some children referred to principles of equal opportunity (e.g., "Boys don't usually get to play with dolls and he wants to try it so he should get a chance like anyone else"). Repeatedly, children referred to principles of fairness (e.g., "Because he never got a chance and it would be mean to not give him a turn"). Children's justifications also demonstrated the understanding and rejection of gender stereotypes. One child explained that exclusion was inappropriate because "boys can play with trucks and girls should be able to also because there is no such thing as boy stuff and girl stuff." These justifica...