Gonzalez, Steele & Baron 2017 paper Flashcards
What is the abstract of this paper?
Studies with adults suggest that implicit preferences favoring White versus Black individuals can be reduced through exposure to positive Black exemplars. However, it remains unclear whether developmental differ- ences exist in the capacity for these biases to be changed. This study included 369 children and examined whether their implicit racial bias would be reduced following exposure to positive Black exemplars. Results showed that children’s implicit pro-White bias was reduced following exposure to positive Black exemplars, but only for older children (Mage = ~10 years). Younger children’s (Mage = ~7 years) implicit bias was not affected by this intervention. These results suggest developmental differences in the malleability of implicit racial biases and point to possible age differences in intervention effectiveness.
What is the rationale behind the importance of this research?
The widespread protests following the deaths of Michael Brown, Eric Gardner, Andrew Loku, Tray- von Martin, Walter Scott, and other unarmed Black men killed by police officers serve as a stark reminder of the continued perception and experi- ence of racial bias in North America. Indeed, research suggests that the media are fraught with negative stereotypes associating Black people with violence, and this biased exposure contributes to ongoing prejudice against Black individuals (Dixon, 2008; Dixon & Linz, 2000; Fujioka, 1999; Mastro & Greenberg, 2000; Tukachinsky, Mastro, & Yarchi, 2015). For example, when non-Black par- ticipants were exposed to news stories about Black criminals, they were more inclined to shoot a Black target, as opposed to White criminals and White targets, in a computerized shooter task (Correll, Park, Judd, & Wittenbrink, 2007). Concep- tually similar studies have shown that stereotypes associating Black people with violence are related to higher rates of shooting Black targets in the shooter task (Correll, Park, Judd, & Wittenbrink, 2002; Correll, Wittenbrink, Park, Judd, & Goyle, 2011; Eberhardt, Goff, Purdie, & Davies, 2004). Together, these studies highlight the relation between media exposure, implicit racial bias, and overt behavior.
Although implicit racial bias is only one of a number of factors that can contribute to the preva- lence of prejudice directed against Black Americans, the broad influence of implicit bias on behavior in a number of domains (e.g., gender, consumer and political preference, alcohol and drug use, etc.; Greenwald, Poehlman, Uhlmann, & Banaji, 2009) underscores the need to understand how best to reduce such deleterious attitudes, with a specific focus on those attitudes that are directly linked to discriminatory behavior toward culturally stigma- tized groups. To date, a number of studies have demonstrated that implicit intergroup biases can be successfully reduced in adults (Lai et al., 2014). For instance, researchers have found that interventions that promote direct (e.g., personal contact with out- group members) and indirect (e.g., reading informa- tion about counterstereotypical exemplars) contact can reduce intergroup bias (Dasgupta & Green- wald, 2001; Foroni & Mayr, 2005; Gonsalkorale, Allen, Sherman, & Klauer, 2010; Pettigrew & Tropp, 2006). Unfortunately, the magnitude of implicit bias reduction is relatively small among adult samples, suggesting that adulthood might not be the optimal period in development to reduce implicit biases (Baron, 2015).
what does this study highlight about implicit bias in childhood?
The results of multiple studies with children sug- gest that implicit preferences for high-status racial groups (e.g., White) over lower status groups (e.g., Black) emerge early in childhood at levels that remain stable across development (Baron, 2015; Baron & Banaji, 2006, 2009; Dunham, Chen, & Banaji, 2013; Newheiser & Olson, 2012; Rutland, Cameron, Milne, & McGeorge, 2005). Furthermore, research also suggests that these biases are stronger in ethnically homogenous communities (McGlothlin & Killen, 2006). However, the developmental invari- ance of implicit biases does not necessarily mean that these biases are equally amenable (or resistant) to change across development (Baron, 2015). Specif- ically, it remains unclear whether developmental differences exist with respect to the capacity to reduce implicit racial bias. In the present study, we examined whether there are developmental differ- ences in the capacity to reduce implicit racial bias through exposure to positive out-group exemplars.
On the one hand, implicit bias might be most amenable to change during early childhood when such cognitions initially form and before they have been extensively reinforced through experience (Devine, 1989; Greenwald & Banaji, 1995; Rudman, 2004). Studies have shown that novel biases, though difficult to reverse, can be reduced significantly immediately following their initial formation (Gregg, Seibt, & Banaji, 2006). There is robust evi- dence that children have acquired implicit racial bias by age 5 (Baron, 2015; Dunham, Baron, & Banaji, 2008), and according to this view, age 5 or earlier might be the optimal period to shape the magnitude of these attitudes.
On the other hand, based on evidence from social cognitive development research, it is plausi- ble that implicit bias might be more amenable to change among older children. Research suggests that cognitive flexibility increases with age (Aboud, 2005; Aboud & Amato, 2001; Bigler & Liben, 2006), and this mechanism might better allow older (as compared to younger) children to shift their evaluations of a racial group after being presented with counterstereotypical information about that group (Lai et al., 2014). Furthermore, in contrast to younger children (5–7 years), children (8–10 years) from majority groups have lower explicit prejudice against out-groups, poten- tially due to motivational processes (Raabe & Beel- mann, 2011). Thus, older children might be more capable of integrating counterstereotypical informa- tion with their prior beliefs about people who are Black. As noted earlier, although implicit biases can be reduced among adults, the magnitude of change is generally quite small. Although cognitive flexibility continues developing into adulthood, implicit biases might be more easily reduced in older children than in their adult counterparts because they have received comparatively less reinforcement of the prevailing cultural attitudes about social groups.
Describe the current study
In the current study we investigated (a) whether children’s implicit racial attitudes can be reduced, and (b) if developmental differences exist in the capacity to reduce implicit racial bias among chil- dren. To address these questions, we adopted an intervention method that has been used successfully to reduce implicit bias among adult populations (Dasgupta & Greenwald, 2001; Foroni & Mayr, 2005; Gonsalkorale et al., 2010; Lai et al., 2014) as well as explicit bias among similarly aged children (Baron, Dunham, Banaji, & Carey, 2014): exposure to counterstereotypical exemplars. Through the use of vignettes, our method exposed participants to Black individuals who were represented in a very positive frame—a depiction that contrasts with the larger cultural messages that contribute to the stigmatization of Black people in North America (Eberhardt et al., 2004; Fujioka, 1999; Mastro & Greenberg, 2000). We used this method to examine whether a brief exposure to vignettes depicting pos- itive Black exemplars, as compared to either posi- tive White exemplars or, in our main control condition, flowers, would reduce implicit racial bias among children.
Our study only included children of Caucasian and Asian ethnicity, two groups that display a clear implicit bias for White individuals over Black indi- viduals (Dunham, Baron, & Banaji, 2006) and who represent the culturally higher status groups in our community. As there is a dearth of research on potential developmental differences in the malleabil- ity of implicit associations, we compared the effec- tiveness of our intervention in younger (early to middle childhood) and older (late childhood to early adolescence) children—groups that differ substan- tially on a variety of measures of cognitive flexibility and executive control (Zelazo, Carlson, & Kesek, 2008). Our study had a 2 (age group: younger or older) 9 3 (condition: Black, flowers, White) 9 2 (ethnic group: Caucasian or Asian) design.
what were the methods of this study?
A total of 369 Caucasian and Asian children between 5 and 12 years (184 boys and 185 girls, Mage = 8.61 years, SD = 1.61) were recruited from a community based science center in April 2013 through September 2014 and were tested onsite in a soundproof room dedicated for behavioral science research. Six of these children were excluded from our analyses due to a profound language barrier (N = 1), experimenter error (N = 2), technical diffi- culties (N = 2), or parent interference (N = 1). An additional 33 children were recruited but excluded as they chose not to complete the study. After these exclusions, as well as four Implicit Association Test (IAT) exclusions detailed in the Results section, our final sample consisted of a total of 359 participants, of whom 257 identified as Caucasian and 102 iden- tified as Asian.
Our testing area at the science center was set up as an exhibit, and parents were able to come into our laboratory space if they were interested in having their child participate in a study. Research assistants also recruited by walking around the science center floor and telling parents about our research. Participants were recruited from a popu- lation with a median income of $75,000. Approxi- mately 85% of parents in this population have received a high school education or higher, and 57% have received a university education or higher. We aimed to collect data from 120 partici- pants in each condition, distributed across the age range. However, due to constraints of our testing location, we ran more children than intended as it is policy at the science center that any child who wishes to participate be allowed to do so (pro- vided parents provide consent). Children who were outside of our a priori range of 5–12 years were excluded from the sample, as were children who were not members of one of the majority (and culturally higher status) groups in Vancouver (either Caucasian or Asian; Statistics Canada, 2006). Parents or a legal guardian of each partici- pant were asked to report their child’s ethnic iden- tification and other demographic information after providing informed consent for their child’s partic- ipation.
frowning face served as category reminders for the attributes good and bad. An image of a White child and an image of a Black child served as category reminders for the two target groups. In front of the monitor on the table were two JellyBeanTM (AbleNet Inc., Roseville, MN, USA) response buttons color matched with the side of the screen in front of which they were placed (yellow on the left, blue on the right). Participants were instructed that any time they saw an image in the middle of the screen or heard a word to determine with which category it belonged (White or Black; good or bad) and to press the corresponding button (yellow or blue) associated with that category.
Participants first began by categorizing faces as either Black or White. In each of these practice trials, a face appeared one at a time in the middle of the screen and participants were instructed to identify as quickly and accurately as possible whether the face was Black or White. There were a total of 12 such practice trials (six Black faces and six White faces). Next, participants practiced classifying words into the categories good and bad. A smiling face and a frowning face were positioned on each reminder bar and for 20 trials participants heard 10 good words and 10 bad words and were similarly asked to cate- gorize them as quickly and accurately as possible.
Following these two practice blocks, children completed a critical test block of trials used to com- pute their association strength. In these trials (N = 30), children used the same buttons to classify an attribute (good or bad) and a target category (Black or White; e.g., Black + good shared one but- ton and White + bad shared another). Children then completed another practice block (N = 20), where they were again asked to classify pictures only, but the sides of the target categories was reversed. In the final test block (N = 30), children again used the same buttons to classify attributes and target categories, but this time, the pairing of the attributes and target categories were switched (e.g., White + good, Black + bad). The side for tar- get categories and attributes were counterbalanced across conditions. This measure is designed to mea- sure the strength of the association between paired stimuli (e.g., Black + good, White + bad, vs. White + good, Black + bad) by recording children’s reaction times during the pairings of stimuli. Specif- ically, we measured the relative positivity and neg- ativity that participants associated with White and Black individuals. For a broader discussion of dif- fering interpretations of response latency tests and of the IAT in particular, see Greenwald, Nosek, and Sriram (2006
what measures were used in this study?
Vignettes
Each participant was read four vignettes. For each vignette in the Black and White conditions, children were introduced to a unique exemplar from that racial group and were told several posi- tive facts about that individual (e.g., “This is James. James lives in North Vancouver where he is a fire fighter. James is an excellent fire fighter and is working hard to become fire chief.”). A photograph of a unique individual in early to middle adult- hood, depending on the story, accompanied each vignette. For two of the vignettes, participants learned about a male exemplar and for the other two vignettes participants learned about a female exemplar. While the pictures of the individuals dif- fered across the Black and White conditions, the descriptions were the same. Each of the four Black and four White target photographs were matched for age and attractiveness. In the main control con- dition, participants heard vignettes about four dif- ferent types of flowers (tulips, daffodils, sunflowers, roses) one at a time. These vignettes included facts about the different flowers as well as positive uses of the flower. These control vignettes also included a photograph of the flower and were designed to match the other two conditions in terms of positive affective valence and length (4–5 sentences; see Appendix S1, for full texts). The order of the vign- ettes was randomly presented.
Child IAT
Children’s implicit racial bias was measured using a child-friendly IAT (Baron & Banaji, 2006). The child IAT measures the strength of association between a target category and an attribute. The two target categories in this test were the racial groups Black and White, and the two attributes were “good” and “bad.” The stimuli used to represent these four categories were the same as reported in Baron and Banaji (2006). The categories Black and White were each represented with four pictures of children from each race matched for age and attractiveness. The “good” and “bad” attributes were presented acoustically. Specifically, children heard four words that could be categorized as “good” (happy, fun, good, nice) or “bad” (yucky, sad, mad, mean). These items were recorded by a woman who spoke each word in an affectively con- gruent manner. On the left- and right-hand side of the screen a yellow (L) and blue (R) reminder bar was present throughout the test. A smiling and a
what was the procedure of this study?
Participants were tested individually. The experi- menter read all instructions aloud to each partici- pant and InquisitTM version 4.0 (Millisecond Software, Seattle, WA, USA) was used to present the study. Participants were randomly assigned to one of three conditions where they heard four posi- tive vignettes. Depending on the condition, these vignettes were about Black individuals, White indi- viduals, or flowers. Children who heard about Black individuals were in our intervention condi- tion, as positive exemplars from this group repre- sent a counterstereotypical portrayal given the broader cultural messages about this group. The flower condition was our main control condition where we presented participants with the same number of vignettes about different flowers designed to induce a positive mood similar to being exposed to positive statements about a person. Because our participants were expected to have a baseline level of implicit pro-White (vs. Black) bias based on prior research with North American sam- ples of these ethnic groups (Baron & Banaji, 2006; Dunham et al., 2006), the condition where children were exposed to positive White exemplars served as an additional control condition as this informa- tion is congruent with broader cultural stereotypes. Following the presentation of the vignettes, partici- pants completed the child IAT (Baron & Banaji, 2006) to measure their implicit attitudes toward the racial categories White and Black.
what were the results of this study?
Prior to conducting our analyses, an IAT score was calculated for each participant using the guide- lines outlined by Greenwald, Nosek, and Banaji, (2003) and Baron and Banaji (2006). This score, called a D score, is a variation of Cohen’s d, and represents the magnitude of a participant’s implicit preference for one group relative to a comparison group. Our data were coded such that a positive score indicated an implicit preference for White individuals over Black individuals, and a negative score indicated an implicit preference for Black indi- viduals over White individuals. Consistent with previous research, children with more than 25% of their response latencies under 300 ms were excluded from the analyses (Baron & Banaji, 2006). Four children were excluded for this reason.
As this is the first study to examine whether there are developmental differences in the malleability of implicit racial bias following an intervention, we performed a median split (Md = 8.39 years), divid- ing our sample into two age groups (a younger group, Mage = 7.36, N = 180, and an older age group, Mage = 9.94, N = 179) that are comparable to age groupings that have been used previously in research on children’s intergroup bias (Baron & Banaji, 2006; Dunham et al., 2006; Raabe & Beel- mann, 2011). Subsequent analyses indicated that results were comparable, even when age groups were divided based on years (ages 5–8 and ages 9– 12; see Appendix S2).
Implicit Racial Bias
To examine whether the magnitude of children’s
implicit racial bias was affected by exposure to pos-
itive Black exemplars, we conducted a 2 (age group:
younger or older) 9 3 (condition: Black, flowers,
White) 9 2 (ethnic group: Caucasian or Asian)
analysis of variance with the IAT D score entered
as the dependent variable. There was no main effect
of age group, F(1, 347) = 0.56, p = .45, g2 1⁄4 :002,
effect of or interactions with ethnic group (Cau- casian or Asian), we collapsed across these groups in subsequent analyses.
In order to examine this interaction, we con-
ducted follow-up analyses of variance for younger
and older children separately. For younger children,
there was no effect of condition, F(2, 177) = 1.26,
p = .29, g2 1⁄4 :01, and post hoc analyses confirmed p
that there were no significant differences between
any of the conditions (ps > .15). As a group,
younger children showed an implicit pro-White
bias (D = .05, SD = 0.18) that was significantly dif-
ferent from chance, t(179) = 3.81, p < .001, Cohen’s
d = 0.57, CI95 = [.02, .08]. For older children, how-
ever, there was a main effect of condition, F(2,
176) = 6.30, p = .002, g2 1⁄4 :07. Post hoc analyses p
revealed that older children in the Black exemplar
condition (D = .01, SD = 0.15) showed significantly
less bias than older children in the flower (D = .09,
SD=0.18; p=.005) or White (D=.11, SD=0.17;
p = .001) control conditions; bias in these control
conditions did not differ (p = .90). Moreover, con-
sistent with the hypothesized effectiveness of the
intervention, after being exposed to positive Black
exemplars, older children’s mean level of bias was
not significantly different from chance, indicating
that they did not show implicit pro-White bias fol-
lowing this intervention, t(61) = 0.43, p = .67,
condition, F(2, 347) = 0.82, p = .44, g2 1⁄4 :005, or p
ethnic group, F(1, 347) = 0.004, p = .95, g2 1⁄4 :001, p
on children’s IAT score. However, there was a sig-
nificant interaction between age group and condi-
consistent with findings from the broader literature on children’s implicit racial bias, older children assigned to the flower, t(58) = 4.05, p < .001, CI95 = [.05, .14], Cohen’s d = 1.06, and White, t
tion, F(2, 347) = 5.92, p = .003, g2 1⁄4 :03 (see
p 95
Figure 1). No other interactions were significant, Fs < 2.30, ps > .10. Because there was no main
control conditions showed an implicit preference for White relative to Black. Simple effects analyses also indicated that there was a significant difference between the younger (D = .07, SD = 0.20) and older (D = .01, SD = 0.15) age groups in the Black condition (p = .03). Older children showed significantly less implicit bias than younger children after being exposed to Black exemplars, suggesting that this intervention was more effective for chil- dren older than age 8.
For comparison, we calculated the effect size of our intervention for older children by comparing the mean of our intervention condition (Black) with the collapsed mean of the two control conditions (flower and White; D = .10, SD = 0.17). For younger children, our manipulation had a very small effect size, t(178) = 0.95, p = .34, CIdiff = [.09, .03], Cohen’s d = 0.15. For older children, our manipula- tion had a moderate effect size, t(177) = 3.53, p = .001, CIdiff = [.04, .14], Cohen’s d = 0.56
what was the discussion of this study?
The results of this study suggest that exposing White and Asian children to counterstereotypical Black exemplars can successfully reduce implicit racial bias among older (Mage = ~10 years), but not younger (Mage = ~7 years), children. Older partici- pants showed an absence of an implicit preference for White relative to Black targets following a brief intervention in which stories about four positive Black exemplars were read. Thus, whereas previous research from the United States, the United King- dom, South Africa, and Canada has found develop- mental invariance in the strength of implicit racial bias as measured by the child IAT among children age 5 and older (Baron & Banaji, 2006; Dunham et al., 2006; Newheiser & Olson, 2012; Rutland et al., 2005; Williams & Steele, 2015), our study sug- gests that important developmental differences exist with respect to the capacity to reduce implicit atti- tudes.
Previous research suggests that adults’ implicit racial biases can be similarly reduced following brief interventions that introduce positive Black exemplars (Cohen’s d = 0.38; Lai et al., 2014). In particular, studies have shown that exposing adults to famous admired Black exemplars and disliked White exemplars reduces implicit pro-White bias (Dasgupta & Greenwald, 2001; Joy-Gaba & Nosek, 2010; Lai et al., 2014). The efficacy of this interven- tion among adults is believed to stem from a shift in the social context that places emphasis on indi- viduals who contrast with usual stereotypes (Lai, Hoffman, & Nosek, 2013). As such, it has been sug- gested that this shift primes subtypes (e.g., “wealthy Black person”; “helpful Black person”) rather than leading adults to revise their prior beliefs about the larger groups (i.e., Black people). Although it is possible that our intervention oper- ates through a similar mechanism for children, it is also possible that older children are instead forming novel associations about this racial out-group. Future research will need to examine whether chil- dren are similarly activating subtypes of their repre- sentations of this racial group or whether they are successfully revising their beliefs about the broader group. Although many studies of attitude change among adults and children speak generally about effective strategies for reducing bias, few studies have identified the conditions under which such change occurs via mechanisms of global attitude change versus the creation of positive subtypes. Future work will need to focus specifically on adju- dicating among these possible mechanisms.
Additional research identifying possible develop- mental differences in the malleability of implicit bias can also shed light on the independent and compet- ing influences of increased cognitive flexibility and longer exposure to cultural messages of bias. Although previous research has found that exposure to positive Black exemplars can similarly decrease the implicit racial bias of adults, we did not include a sample of adults in our current study, and thus we are unable to determine whether implicit racial atti- tudes might actually be more malleable among older children relative to adults. An interesting possibility to consider is whether there is a curvilinear relation between age and the magnitude of attitude change following similar interventions, with older children representing an age group that has comparatively less reinforcement of biases but enough cognitive flexibil- ity to overcome the initial bias that has formed. This would suggest that late childhood might be a particu- larly effective time for interventions designed to reduce children’s implicit racial biases.
Interestingly, our intervention did not success- fully reduce bias in younger children, a finding that highlights possible developmental differences in the malleability of implicit associations. As discussed previously, research provides evidence that older children have more cognitive flexibility, which might allow them to more easily alter their existing beliefs about social groups (Aboud, 2005; Aboud & Amato, 2001; Bigler & Liben, 2006). An alternative possibility to consider is that our intervention was more effective with older children because of the particular strategies involved. Specifically, for our intervention to be effective, children needed to cate- gorize the individuals in our vignettes as members of a particular racial category and generalize that affective association with novel members of the cat- egory. Research suggests that young children may not spontaneously categorize others by race to the same extent as older children. Even though these children can sort faces by race, they may be less likely than older children to spontaneously attend to race when reasoning about others (Pauker, Wil- liams, & Steele, in press). Moreover, research sug- gests that younger children often privilege other categories (e.g., gender, language) over race on a variety of reasoning tasks (Degner & Wentura, 2010; Kinzler, Shutts, DeJesus, & Spelke, 2009; Shutts, Banaji, & Spelke, 2010). It will therefore be important to examine other intervention strategies with younger children in order to identify why young children’s implicit racial attitudes might be more resistant to this type of intervention and to ensure that their lack of implicit attitude change in the present study was not due to idiosyncratic aspects of our manipulation.
Additional research would also help to identify whether developmental differences exist in the effi- cacy of various intervention strategies in creating long-term attitude change. Although a number of studies with adults have investigated short-term atti- tude change, the duration of these interventions has not yet been fully assessed (Lai et al., 2014). Examin- ing longer term effects of this and other intervention strategies can help to reveal whether children’s implicit attitudes have genuinely shifted or whether the particular intervention strategy employed (e.g., exposure to counterstereotypical exemplars) is effec- tive only in the short term. Although our findings present a promising method of implicit racial bias reduction in children, an important consideration for future research is that developmental differences must be considered when designing and employing interventions to reduce implicit bias. The current study, coupled with future work on developmental differences in reducing implicit racial bias, can help to determine whether there is an optimal period in development for targeted interventions to decrease bias, and importantly, what types of interventions are most effective at each age.
