race, social class, and student engagement in middle school english classrooms

15
Race, social class, and student engagement in middle school English classrooms q Sean Kelly University of Notre Dame, Center for Research on Educational Opportunity, 1015 Flanner Hall, Notre Dame, IN 46556-5611, USA Available online 18 September 2007 Abstract Student level data on participation in classroom discourse and student effort on assignments in 117 middle school Eng- lish classrooms are used to investigate the social determinants of student engagement in classroom instruction. Social iden- tity theories of race, social class, and attachment to school, and research in the social psychology of achievement motivation both suggest differential levels of student engagement among diverse student groups. Using multilevel models, the author investigates the relationship between classroom context and students’ levels of engagement. Levels of engage- ment among black and low SES students are mostly insensitive to classroom context, suggesting there is little collective action directed at fostering anti-school norms among these student groups. However, consistent with research in the social psychology of achievement motivation, students who begin class with weaker reading and writing skills are less likely to be engaged, setting the stage for a cycle of reduced achievement growth. Ó 2007 Elsevier Inc. All rights reserved. Keywords: Educational inequality; Student engagement; Social identity theory; Achievement motivation; Literacy 1. Introduction In the sociology of education, there is a rich theoretical tradition which posits that students’ race and social class identities work in powerful ways to affect their attachment to school in the broadest sense, and their level of classroom engagement on a day to day basis. Social identity theories of race, social class, and attachment to school stress how group membership and peer group interactions affect student engagement. Research in the social psychology of achievement motivation also predicts that minorities and disadvantaged students will have lower levels of engagement, but the social-psychological process of withdraw from school is linked to the individual’s response to the competitive environment of schools rather than intra-group behavior. The concept of student engagement (Finn, 1989; Newmann, 1992) offers a way to link the observed inequal- ity in achievement growth among social groups with the larger social forces that act on students of color, and 0049-089X/$ - see front matter Ó 2007 Elsevier Inc. All rights reserved. doi:10.1016/j.ssresearch.2007.08.003 q Paper prepared for the annual meeting of the American Educational Research Association, April, 2004, San Diego, CA. The author would like to thank Adam Gamoran and Bill Carbonaro for their helpful comments. The Partnership for Literacy Study was funded by the U.S. Office of Educational Research and Improvement. Additional funding for this analysis was provided by The Spencer Foundation and the College of Arts and Sciences at the University of Wisconsin-Madison. E-mail address: [email protected] Available online at www.sciencedirect.com Social Science Research 37 (2008) 434–448 www.elsevier.com/locate/ssresearch

Upload: sean-kelly

Post on 12-Sep-2016

219 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Race, social class, and student engagement in middle school English classrooms

Available online at www.sciencedirect.com

Social Science Research 37 (2008) 434–448

www.elsevier.com/locate/ssresearch

Race, social class, and student engagement in middleschool English classrooms q

Sean Kelly

University of Notre Dame, Center for Research on Educational Opportunity, 1015 Flanner Hall, Notre Dame, IN 46556-5611, USA

Available online 18 September 2007

Abstract

Student level data on participation in classroom discourse and student effort on assignments in 117 middle school Eng-lish classrooms are used to investigate the social determinants of student engagement in classroom instruction. Social iden-tity theories of race, social class, and attachment to school, and research in the social psychology of achievementmotivation both suggest differential levels of student engagement among diverse student groups. Using multilevel models,the author investigates the relationship between classroom context and students’ levels of engagement. Levels of engage-ment among black and low SES students are mostly insensitive to classroom context, suggesting there is little collectiveaction directed at fostering anti-school norms among these student groups. However, consistent with research in the socialpsychology of achievement motivation, students who begin class with weaker reading and writing skills are less likely to beengaged, setting the stage for a cycle of reduced achievement growth.� 2007 Elsevier Inc. All rights reserved.

Keywords: Educational inequality; Student engagement; Social identity theory; Achievement motivation; Literacy

1. Introduction

In the sociology of education, there is a rich theoretical tradition which posits that students’ race and socialclass identities work in powerful ways to affect their attachment to school in the broadest sense, and their levelof classroom engagement on a day to day basis. Social identity theories of race, social class, and attachment toschool stress how group membership and peer group interactions affect student engagement. Research in thesocial psychology of achievement motivation also predicts that minorities and disadvantaged students willhave lower levels of engagement, but the social-psychological process of withdraw from school is linked tothe individual’s response to the competitive environment of schools rather than intra-group behavior.

The concept of student engagement (Finn, 1989; Newmann, 1992) offers a way to link the observed inequal-ity in achievement growth among social groups with the larger social forces that act on students of color, and

0049-089X/$ - see front matter � 2007 Elsevier Inc. All rights reserved.doi:10.1016/j.ssresearch.2007.08.003

q Paper prepared for the annual meeting of the American Educational Research Association, April, 2004, San Diego, CA. The authorwould like to thank Adam Gamoran and Bill Carbonaro for their helpful comments. The Partnership for Literacy Study was funded bythe U.S. Office of Educational Research and Improvement. Additional funding for this analysis was provided by The Spencer Foundationand the College of Arts and Sciences at the University of Wisconsin-Madison.

E-mail address: [email protected]

Page 2: Race, social class, and student engagement in middle school English classrooms

S. Kelly / Social Science Research 37 (2008) 434–448 435

students from disadvantaged backgrounds. However, there is conflicting evidence on the level of engagementof minority and disadvantaged students relative to white/privileged students, and contrasting social-psycho-logical processes have been identified which alternately foster or hinder engagement among diverse students.In this analysis, I reconsider the relationship between race, social class, and student engagement in a diverse setof classrooms. Unlike much of the prior research on student engagement, this analysis uses unique observa-tional data on classroom participation in middle school English classrooms. I find little evidence to supportsocial identity theories of reduced engagement among minority and disadvantaged students. In contrast, stu-dents who begin class with weaker reading and writing skills are less likely to participate in classroom dis-course and exert effort on classroom tasks, setting the stage for a cycle of reduced achievement growth.

2. Two theories of attachment to school

2.1. Social identity theory

Social identity theory was developed by European social psychologists in the 1970s as a theory of inter-group conflict (Tajfel and Turner, 1979 [1986]). The basis of social identity theory is that individuals striveto maintain a positive social identity, an idea which is central to sociologist’s conception of the self (Goffman,1963). Tajfel and Turner extended this logic to groups. Social identity theory states that if social groups or cat-

egories are differentiated along a status dimension, then low status group members will be driven to correct their

low status, either individually or collectively. Tajfel and his colleagues developed social identity theory fromempirical findings in laboratory settings, but educational researchers also advanced social identity theoriesof group behavior based on observational research in schools. Importantly, close observations of schoolingrevealed that the motive to maintain a positive social identity can lead to withdraw from school, and peer-group behavior which supports anti-school norms.

Ethnographic case studies of working class youth by Willis (1977 [1981]), McLaren (1986), Eckert (1989),and others, depict students with highly oppositional alignments to school. The ‘‘Lads,’’ ‘‘Burnouts,’’ and otherpeer-groups eschew the ideology of school achievement, rejecting the notion that an education providesopportunity. Low-achieving and marginalized within the schools they attend, these peer-groups maintain apositive social identity not by challenging their low status position vis-a-vie the school, but by engaging ina process that might be termed social creativity. A positive social identity is achieved by creating valued alter-natives to school achievement, which might be working on cars, being accomplished on the soccer pitch, orbeing a sought after date on the weekends (for the Lads it seems to have been a stripped-down alternativecentered around hell-raising and exhibiting a particular type of ‘‘cool’’). Developing alternative sources of suc-cess and a positive identity—and really believing in them—requires collective effort, and is undermined byindividuals who conform to the school’s definition of success. Thus, an important component of collectivelyrejecting school is monitoring and sanctioning any peer-group members who do show interest in school. Theresulting anti-school attitudes and behavior permeate every aspect of schooling, from classroom behavior, tofield trips, to after-school forays in vandalism and substance abuse.

In a series of books and articles beginning in the 1970s, John Ogbu and colleagues have argued that blackstudents are also prone to developing anti-school attitudes. The relative disadvantage of blacks, coupled withtheir racial identity, leads to reduced attachment to school (Ogbu, 1990; Fordham and Ogbu, 1986). Compar-isons made between blacks, who are involuntary minorities, and voluntary minority groups, like the PunjabiSikhs (Gibson, 1987) highlight black American’s oppositional frame of reference. An important element ofthis multifaceted argument highlights peer-group norms and interactions. Ogbu theorized that black youthsanction peers who show interest in school by labeling school attachment as ‘‘white.’’ The ‘‘Burden of ActingWhite’’ is a social-identity theory of reduced student engagement similar to the research on peer-group behav-ior among working-class youth but focusing on racial identity instead.1

1 The hypothesis that rejecting school in favor of other goals works to reduce status anxiety is not unique to research on social class orrace, it can also be found in the multitude of research on differentiation–polarization theory (see Berends, 1995).

Page 3: Race, social class, and student engagement in middle school English classrooms

436 S. Kelly / Social Science Research 37 (2008) 434–448

Ogbu’s theory is certainly plausible; particularly because racial group identities are highly salient. But themere fact that there is a burden of acting white among some black students reveals that they can and do focuson individual mobility rather than social creativity. Moreover, a sizeable body of research now exists whichsuggests that among black students, peer group processes that create a burden of acting white are eithernot widespread, or not very powerful. Early research on the attitudes of blacks toward schooling found thatin fact blacks have not lost faith in education as a means to social mobility (Patchen, 1982). Mickelson (1990)began to explain the paradox between faith in schooling and low achievement by noting that concrete attitudestoward school, those more closely related to a black student’s particular family and peer influences, were muchmore negative than their abstract attitudes. Evidence from recent large-scale quantitative research shows thatblack students are not in fact any more likely to have anti-school attitudes than whites (Morgan and Mehta,2004; Ainsworth-Darnell and Downey, 1998; Cook and Ludwig, 1997), although teachers do perceive blackstudents as putting forth less effort (Ainsworth-Darnell and Downey, 1998). There is also indirect evidencethat among black peer groups, high performing students are not sanctioned as Ogbu’s theory contends (Ains-worth-Darnell and Downey, 1998; Cook and Ludwig, 1997; Roscigno, 1998; Datnow and Cooper, 1997).Quantitative research provides little support for the notion that black students are less attached to school,let alone that there is a burden of acting white.

Several qualitative studies of peer-group interactions among blacks illustrates why many black studentsmaintain a positive overall approach to schooling, despite the fact that they must overcome negative teacherperceptions of behavior (Tyson, 2003), low-track placements (Gamoran and Mare, 1989), and other barriersto success. MacLeod (1987) attributed the pro-school attitudes of the ‘‘Brothers,’’ an impoverished African-American peer-group, to a form of social creativity social identity theorists call changing the out-group with

which they compare themselves. MacLeod found that these youth compared themselves not with whites,and other more advantaged groups, but with their parent’s generation. Even though their chances of success(and current status) were low relative to advantaged white students, they pursued success in school because itwas clear to them that their chances of success were considerably greater than that of their parents’ generation.While not denying the existence of negative peer pressure among black students, Horvat and Lewis (2003)found that it was offset by equally strong peer networks supportive of school success. Carter (2005) found thatwhile minorities do sometimes label behavior as ‘‘acting white’’ in an effort to foster solidarity and culturalpride, this is not primarily a distinction attached to academic behaviors; the goal seldom appears to be to sanc-tion someone with educational aspirations. Similarly, Tyson et al. (2005) reported that disparagement of high-achieving students was not particularly racialized; only a small fraction of black students were targeted as‘‘acting white,’’ and there was little effect of this treatment on their course enrollments.

Both race and social class-based theories of resistance apply social identity theory to understanding socialinequalities in attachment to school.2 These theories seem convincing because they are integrated into a larger,well supported theoretical framework in the social psychology of intergroup conflict. Moreover, the portraits ofsocial creativity in action are compelling, because the effects appear to be so powerful when in place, amountingto the complete rejection of any academic achievement ideology. But is this type of collective action wide-spread? Education does in fact provide an opportunity for social mobility and many students appreciate thatopportunity. Unfortunately, the experimental research on intergroup conflict provides little evidence on theconditions that make social creativity likely, on when or how often it should be expected. In this research I revi-sit these theories, considering not attachments to school in general, but measures of within-class studentengagement. Because students have intersecting social identities, with a disproportionate number of black stu-dents being from lower social class backgrounds, race and social class must be examined simultaneously.

Even when taken together though, social identity theories of reduced engagement are incomplete, becausestudents from different race/ethnic and social class backgrounds also differ in the skills they bring to class, ontheir levels of initial achievement. If students begin school with lower levels of achievement, then we mightexpect these students to become less engaged, regardless of group membership or the actions of their peers.

2 Different terminology is sometimes used to describe these theories and to put them with other theorists into categories such as ‘‘critical’’theory, and ‘‘resistance’’ theory (McFarland, 2001). Willis himself used the term, ‘‘cultural production,’’ to denote its connection withother social reproduction theories, while Ogbu attached different terms to specific parts of the theorized social process (i.e., ‘‘oppositionalframe of reference’’ and ‘‘burden of acting-white’’).

Page 4: Race, social class, and student engagement in middle school English classrooms

S. Kelly / Social Science Research 37 (2008) 434–448 437

2.2. Theories of achievement motivation

Self-worth (Covington and Berry, 1976) and Self-efficacy (Bandura, 1977) based explanations of achieve-ment motivation stress that perceptions of ability are an important determinant of student engagement, inde-pendent of other social identities or peer group processes. This explanation explains engagement in the contextof socially risky tasks, where one’s performance is evaluated relative to others. Thus, it is an attractive theoryfor explaining effort exerted in the classroom, because classrooms are rich in evaluation, the basis of which issocial comparisons among classmates. Students with a low sense of self-worth or efficacy on a task are morelikely to engage in failure-avoiding behaviors, in particular, avoiding effort. When little effort is exerted onecan maintain the belief that one could potentially succeed on the task, or at least, no evidence to the contraryis publicly presented. Students who enter class with somewhat weaker skills than their peers will have an incen-tive, at least in the short-term, not to participate in socially risky tasks which would further solidify their iden-tity as a poor student. Over-time a reciprocal process of disengagement, further deterioration in levels of self-efficacy and self-worth, and reduced achievement growth unfolds.

Consistent with this scenario, research has shown that (a) even very young children utilize social compar-ison to evaluate their ability levels (Smith et al., 1987), (b) that over time, negative evaluations become morecommon (Ruble and Frey, 1987), and (c) that low-achieving students are much less likely to be engaged thanhigh-achieving students (Shernoff et al., 2003; Roeser et al., 2002; Valeski and Stipek, 2001; Rosenbaum, 2001;Roeser et al., 2000; Ainsworth-Darnell and Downey, 1998; Voelkl, 1997; Finn et al., 1995). A few studies havelooked specifically at participation in classroom discourse, one of the measures of engagement used in thepresent analysis. Morine-Dershimer (1985) found that high achieving students participate more in classroomdiscourse and public tasks (e.g., going to the chalkboard) than low achieving students, as do students whoscore high on teacher reported measures of ‘‘academic status.’’ By comparing students across grade levels,Good et al. (1987) concluded that lower achieving students become more passive learners over time, askingfewer questions.

3. Social inequalities in student engagement: a multilevel perspective

My goal is to explore the sources of differential student engagement among diverse students. Learningoccurs within the social setting of the classroom, and it is necessary to consider how the composition of a classaffects the learning environments of students from different backgrounds. In order to understand the sourcesof differential student engagement in middle school English classrooms I employ a multilevel perspective (Brykand Raudenbush, 1992 [2002]). An individual student’s level of engagement can be described by the meanengagement of his or her class, along with the student’s deviation from that mean, the level of engagementrelative to other students within the class.

Across classrooms on average, initial achievement, the skills and knowledge students begin the school yearwith, is likely to be an important determinant of student engagement within classrooms. As students seek tomaintain a positive sense of self-worth, the variation in achievement and perceived competence will be asso-ciated with varying levels of engagement. When individuals within a class are similar on initial achievement,social comparisons may result in fewer negative self-evaluations, so levels of engagement may be more even inclassrooms with little variation in initial achievement. In practice though, I expect to find initial achievementeffects in almost every class, because all of the classrooms in this study contain students with a range ofachievement levels, and relatively small differences in achievement could still produce problems of achievementmotivation. Another source of variation in the achievement-engagement relationship across classrooms is thelearning environment itself. Applied research on instruction in heterogeneous classrooms shows that the rela-tionship between initial achievement and engagement is linked with the nature of evaluation across classrooms(Cohen and Lotan, 1997).

Because race and social class are associated with levels of achievement, much of the differences in engage-ment amongst students associated with these factors may be explained by the achievement-engagement con-nection. In other words, much of the race and social class gaps in engagement may have little to do with raceand social class per se. Residual differences among students, after controlling for initial levels of achievement,represents the potential influence of social creativity processes on engagement among black students and stu-

Page 5: Race, social class, and student engagement in middle school English classrooms

Table 1Characteristics of the partnership for literacy classrooms

0–25% 26–50% 51–75% 76–100% Total

Percent black 82 (70%) 11 (9%) 3 (3%) 21 (18%) 117Parent’s highest education (%) college graduates 0 18 (15%) 72 (62%) 27 (23%) 117

1–10 11–15 16–20 21+ TotalClass size (# of participating students) 9 (7%) 31 (26%) 43 (37%) 34 (29%) 117

6th 7th 8th TotalGrade level 1 (1%) 57 (49%) 59 (50%) 117

N = 117.

438 S. Kelly / Social Science Research 37 (2008) 434–448

dents of lower social class. But the remaining differences could also be due to common cultural sources, orother common factors, rather than collective action.

How can a social creativity effect be more effectively isolated? Research on social solidarity among groupsprovides a potential answer. As a form of collective action, social creativity will be more effective when mon-itoring of adherence to norms of social creativity and sanctioning of those who violate the norms, are easier(Hechter, 1987). For example, in order for black students to be deterred from pursuing academics by beinglabeled white, pro-school behaviors must be readily identified by peers and publicly sanctioned. I hypothesizethat the effects of social creativity should be strongest in classroom settings where more of the student’s groupmembers are present, because monitoring and sanctioning of group members will be strongest in these set-tings. This conception of collective action posits a multi-level interaction between the social composition ofclassrooms, and the levels of engagement of students within that classroom relative to one another. For exam-ple, when a class becomes predominantly low-SES, levels of engagement among low-SES students will be fur-ther reduced compared to any high-SES students.

3.1. Summary

Research on the dynamics of race and social class in schools suggests that rather than low-status studentsusing school as a vehicle for social mobility, eventually casting off their negative social identities, they may usesocial creativity to reduce the negative effects of being members of a group that traditionally performs poorlyin school. While this social creativity benefits students cognitively in the short-run, it does little for their actualsuccess in school, and all that this will mean to them over their life course. How important is this process ofsocial creativity; to what extent can it account for social inequalities in engagement? Prior research has failedto simultaneously control for a potentially confounding variable, the student’s level of initial achievement.This omission is striking given the fact that educational researchers find such strong links between achieve-ment and/or perceived competence and engagement (Shernoff et al., 2003; Finn et al., 1995). This analysis con-siders the potential sources of student engagement stemming from levels of achievement and race/social classidentities simultaneously using observational data from a large array of classrooms.

4. Data and methods

This study uses data collected by the National Research Center on English Learning and Achievement(CELA). The Partnership for Literacy study collected data on two cohorts of teachers in Wisconsin andNew York state over a two year period. A total of 63 classrooms located in 19 schools participated in the datacollection in 2001–2002, while 57 classrooms located in 23 schools participated in 2002–2003. The participa-tion rate for students across all classrooms was 82.8%, leaving 2051 students in 117 classrooms. Most analysesrely on 1,922 students with fall achievement data.3

3 A little less than half of the non-participating students refused to participate, the remainder failed to turn in parent or studentpermission forms. Partnership classrooms contained 1149 participating students in 2001–2002, and 1228 participating students in 2002–2003. Of these 2377 students, 2105 returned student questionnaires. Twenty-two students were missing all data on race/ethnicity or familybackground, and three classrooms had to be dropped because of a lack of observational data.

Page 6: Race, social class, and student engagement in middle school English classrooms

S. Kelly / Social Science Research 37 (2008) 434–448 439

The Partnership for Literacy study contained only regular-track classrooms, or untracked (heterogenousability-group) classrooms. Table 1 displays some other important characteristics of the partnership class-rooms. The classrooms are evenly split between 7th and 8th grade classrooms, and contain a range of classsizes. Because of the schools sampled, many of the classrooms are rather racially homogenous, a large numberhave predominantly white or black student populations. More classrooms are heterogeneous with regard tofamily background. For example, about 81% of the variance in parent’s occupational education lies withinclassrooms (Appendix A provides full descriptive data for the Partnership classrooms).

4.1. Measuring student engagement in English and language arts classrooms

This analysis draws on two sets of student engagement measures, student-reported measures of effort, andobservational data on participation in classroom discourse. Partnership classrooms were observed twice eachin the fall and spring on adjacent days when teachers were conducting literature lessons. Previous CELA anal-yses (Nystrand, 1997; Applebee et al., 2003) revealed that properties of classroom discourse could be reliablyestimated using four observations. The reliability of question answering and asking counts at the student levelaveraged about.80 in the present study.4 Researchers videotaped each classroom observation and recordedinstructional properties using the CLASS software (CLASS 4.24), a real-time, laptop-based data collectionsystem. Files were later edited in the laboratory using the videotapes to improve coding consistency. Amongother instructional properties, CLASS records properties of classroom discourse during question and answersessions associated with student engagement including:

Source of Question (Teacher or Student).Response (was the question answered; if by a student, who was the student; were there multiple responses).Authenticity (was the question a test question, in other words, did it have a predetermined answer, or was itauthentic).Uptake (did the question incorporate a student’s previous response, if so was the uptake genuine uptake, orwas it test uptake, with the student’s previous response entirely predictable).Cognitive Level (Reporting vs. Thinking. Cognitive level distinguishes recitations or reports of what hap-pened, with unknown information such as an analysis, generalization, or speculation based on the text).Nature of response (Did the respondent, typically the teacher, answer the question, close the question downand move on to another question, or open the question up for others to answer; if answered, was the ques-tion answer simply, or was it elaborated with a rationale given).

Not all participation in classroom discourse indicates a similar level of engagement. Throughout the anal-ysis, I distinguish between procedural participation and substantive participation in classroom discourse(Nystrand and Gamoran, 1991). Procedural participation is illustrated by asking or answering non-authentic,low cognitive level test questions, while substantive participation is illustrated by asking or answering eitherauthentic questions, or questions that involve high level thinking, or posing a question with authentic uptake.When students are procedurally engaged they can get by without much critical thinking, or being personallyconnected to the topics covered. Classrooms where teachers facilitate substantive engagement by posingauthentic questions, and questions with uptake, and where teacher-directed lessons often give way to discus-sion are called ‘‘dialogic’’ classrooms (Nystrand, 1997; Applebee et al., 2003).

Over the two year period of data collection, 29,673 questions were recorded by the class program duringquestion and answer sessions where the teacher directed the flow of the instruction. Of these, 12,322 couldnot be associated with individual students, either because the class answered by a show of hands, in chorus,or the student could not be identified. Of the 17,351 questions identified with individual students, 15,575 wereasked by the teacher and responded to by a student, and 1776 were student questions.

4 The reliability of question answering and asking counts aggregated to the student level was .83 for procedural engagement, and .77 forsubstantive engagement. In a previous CELA study, perfect agreement between raters on specific question properties ranged from a low of78% for authenticity, to 81.7% for uptake (Nystrand, 1997).

Page 7: Race, social class, and student engagement in middle school English classrooms

440 S. Kelly / Social Science Research 37 (2008) 434–448

In addition to the main set of question-level observational data on student engagement, five measures ofstudent effort in English and language arts were included on the year-end student questionnaire. In separatequestions students were asked how often they completed reading and writing assignments (on 7 point scalesfrom never to every time). On the same scale, they were asked how often they revise or rewrite papers, andgiven another chance to reiterate that they never did so, on a follow-up question. Finally, they were askedhow much time they spend on English homework each week (on a 7 point scale from none to four hoursor more). These measures were used to create a scale of student effort (a = .57).

4.2. Measuring race, social class, and initial student achievement

4.2.1. Race/ethnicity

On the student questionnaire individuals were asked to report whether or not they were white, black or African-American, American-Indian, Asian or Pacific Islander, or some other race, and whether or not they were Hispanic.Students were allowed to put themselves in more than one category, and many did so (e.g., 41 students reportedbeing both white and black). In this study many of the Asian-American students were children of Hmong refugees.

4.2.2. Social class

In this analysis, I take the perspective that a student’s social class identity is multi-faceted, and I use twomeasures of status attainment taken from the student questionnaire, the educational and occupational attain-ment of the student’s parents, and a measure of family resources.

The student was first asked to report the highest educational attainment of his or her mother/step-motherand father/step-father or guardians on a five category scale (<high school, high school graduation only, somecollege, college graduate, graduate or professional degree). Because of the high number of students living insingle-family homes, this analysis combines the education of both parents into a single measure, parent’s high-est educational attainment. Students were then asked to list their parents’ jobs and describe what they do atwork. Using this data, two dimensions of occupational attainment (occupational education and earnings) werecoded using Hauser and Warren’s (1997) update of socio-economic indices for the 1990s.

These measures describe the general process of status attainment, with parents obtaining an education, andthen obtaining jobs in occupations that vary on average levels of education and earnings. The correlationbetween occupational education and occupational income in these data is .817, and parent’s educational attain-ment is correlated.53 and.6 with the occupational income and education measures respectively. The third mea-sure of social class, the family resources scale, is taken from a question asking students to list if they havecertain resources in the home. The list contains 11 items, such as a VCR, an electric dishwasher, a specific placefor study, a computer, etc. When used with the measures of status attainment, the family resources scale mea-sures the extent of resources, educational or otherwise, that the family is able to provide for their children. Thefamily resources scale has a correlation of about.4 with each of the individual education and occupationalattainment measures. An additive scale of the social class variables is used in this analysis (family resources,occupational education, occupational income, and parent’s education; a = .77).

4.2.3. Initial achievement

At the start of the year CELA researchers administered two assignments, a literature assignment askingthem to respond to two poems by author Gary Soto, ‘‘Oranges’’ and ‘‘Finding a Lucky Number,’’ and a writingassignment about an admirable character from a movie or book. The reading comprehension and writingassignments were combined to create a fall reading achievement scale (a reliability = .63).5 Further informa-tion about the CELA-administered assignments is provided in Langer et al. (2005).

5 Spring assignments were also administered. Fifteen cases were removed from the analysis as outliers because their yearly gain scoreswere abnormally high (24+ point gain where scores ranged from 0 to 28) or low (7+ point loss) suggesting a potential measurement erroron the fall exam. The analysis was also conducted on a subset of 592 students in Wisconsin with standardized achievement data collectedby the school districts. Those results reinforce the results presented here.

Page 8: Race, social class, and student engagement in middle school English classrooms

Table 2Participation as a function of race, social class, and CELA achievement score

Procedural participation Substantive participationa

1 2 3 4

Within-class

Male .29 (.049)*** .39 (.050)*** .20 (.047)*** .27 (.048)***

Hispanic �.12 (.087) �.027 (.086) .057 (.51) .11 (.086)American Indian .044 (.10) .15 (.10) �.17 (.11) �.14 (.11)Asian �.40 (.15)** �.38 (.15)* �.26 (.14) �.27 (.14)Black �.017 (.086) .11 (.086) .19 (.085)* .25 (.085)**

SES scale .19 (.034)*** .16 (.034)*** .19 (.033)*** .18 (.033)***

CELA achievement scale .23 (.037)*** .19 (.033)***

Random effect (SD) .13 .063Variance explained (%) 7.6% 11.8% 5.2% 7.3%

Between-class

Percent black �.78 (.31)* �.58 (.32) �.38 (.25) �.38 (.26)SES scale �.88 (.28)** �1.05 (.30)*** .17 (.22) .22 (.24)CELA achievement scale .36 (.21) �.030 (.17)Variance explained (%) 4.7% 6.6% 16.0% 14.0%

Poisson models, variable exposure, adjusted for over-dispersion. 1922 students in 117 classrooms.a Year of data, experimental cohort of classroom, and year by cohort interaction at the school level added to models of substantive

participation. Coefficients not reported.* P < .05.

** P < .01.*** P < .001.

S. Kelly / Social Science Research 37 (2008) 434–448 441

4.3. Quantitative models of student engagement

The data are modeled using multi-level regression models for count outcomes. Students are nested withinclassrooms, and the multilevel analysis partitions the variance in the dependent variable into between-andwithin-class components. At the within-class level, the estimated coefficients indicate differences in questionasking and answering relative to other students within the class. Differences across classrooms in average levelsof participation are captured by a class-level intercept. The primary variables of interest at the between-classlevel are compositional variables (like the percentage of low-SES students in a class). At the between-classlevel, controls are added when necessary to control for the study design of the Partnership for Literacy(cohort, year, and cohort · year interactions).

Poisson regression models estimate counts of substantive and procedural question asking and answeringacross all classroom observations, with the models adjusted for variable exposure. Poisson regression modelsestimate expected counts as a multiplicative function of an incidence rate and an exposure. In the case of thepartnership, variable exposure occurs at the between-class levels, as some classes were observed three, or fivetimes instead of four, and some classes spent more time in Q&A. All of the models adjust for the variableexposure due to the number of observations, so that the class-level intercept reflects differences in the quantityof time spent in Q&A of different types of classes, holding the number of observations constant.6 The analysesof student effort have a similar set of predictors and multi-level structure as the models of participation, butthe outcome is a linear variable rather than a count outcome.

6 Unfortunately, attendance data was not recorded for individual students, so variable exposure within classes cannot be accounted for.HLM5’s (SSI, 2000) over-dispersion adjustment is further used to correct for the distribution of the count outcome. Generating an over-dispersion parameter relaxes the Poisson regression assumption that the conditional variance is equal to the conditional mean. In thesedata, the unconditional variance in the dependent variable is greater than the unconditional mean, so this is a desirable procedure. Inpractice though, adding the over-dispersion parameter does little to affect the estimated coefficients and standard errors, perhaps becausethe models already take into account important predictors of the heterogeneity in classroom participation.

Page 9: Race, social class, and student engagement in middle school English classrooms

442 S. Kelly / Social Science Research 37 (2008) 434–448

5. Results: the social distribution of participation in classroom discourse and student effort

5.1. Participation in classroom discourse

Unconditional models reveal that about 3/4 of the variance in question asking and answering is betweenstudents in the same classroom.7 The number of questions a particular student answers or asks depends pri-marily on which individual student is being considered, and to a lesser, but still significant extent on the classthat particular student attends. The correlation between counts of procedural and substantive participation is.433 at the student level, and .284 at the class level.

5.1.1. The within-class effects of race and social class

Models 1 and 3 in Table 2 describe the racial and social class distributions of student participation. Unlessotherwise stated, all within-class variables are group-mean centered; thus they report a pooled comparison ofblacks to whites within the same classrooms. Whether one considers procedural or substantive participation,within the same classrooms, black and white students participate in classroom discourse at about the same level.Higher SES students participate more frequently than do lower SES students. A one standard deviation increasein social class results in a 1.18 factor increase in instances of substantive participation, a little less than the effectsize of the difference between boys and girls (1.22 in Model 3). In addition to the social class and gender differenceswithin the Partnership classrooms, there is also a negative effect of being Asian-American.

Table 3 provides a comparison of question asking and answering among high and low SES students, describ-ing the baseline differences in more detail. These social class comparisons provide a good example of the size ofsome of the within-class disparities in participation. Despite the low percentage of variance explained in Models 1and 3, the effect of social class on participation is dramatic. Within the same classrooms, high SES students bothask and answer questions about 30% more than low SES students. High SES students are even more likely toanswer authentic questions, and questions with authentic uptake. The following exchange during a questionand answer session on The Outsiders gives an example of authentic uptake, where the teacher moves off script,putting a student response at the center of the classroom discourse, urging the class to respond to John’s question.

Teacher: What made the Socs unhappy?Student: I think they were mad because the Greasers were happy.Teacher: Ok, that’s an interesting thought. Does anybody want to respond to that? John said, ‘‘I think theSocs were mad because the Greasers were happy.’’

These teacher questions also have a high cognitive level; they ask the students to provide an analysis of one ofthe gangs’ character traits. High SES students were about 42% more likely to ask and answer questions thatprompted generalization or analysis, as opposed to a simple report. High SES students appear to participatemore in almost every type of classroom discourse, but the differences are greatest in more substantively engagingforms of discourse.

Table 2 also contains estimates of differences in the quantity of time spent in the two types of Q&A inclasses with different race and social class compositions (the between-class coefficients). At the class level,predominantly black and low-SES classrooms have lower levels of procedural discourse. Because mostquestions asked during Q&A sessions are non-substantive ‘‘test’’ type questions I refer to in this analysisas procedural, we can say that there is simply somewhat less structured talk among teachers and students in

general in predominantly black and low-SES classrooms.8 Under this specification, with group-mean cen-tered race variables, the class level coefficient for racial composition tests whether all students, including

7 28.6% and 24.3% of the variance in procedural and substantive participation in Q&A lies between classes, respectively.8 Predominantly black classrooms also engage in less substantive Q&A in reduced form models (without SES). What type of instruction

occurs in predominantly black classrooms if less time is spent in question and answer sessions? Predominantly black classrooms spendabout the same amount of time engaged in class discussions and lectures as white classrooms, and procedures and directions, which allclassrooms devote a substantial amount of time to. Predominantly black classrooms spend less time doing small group work and silentreading, and more time is spent dealing with disciplinary actions, interruptions, and reading aloud and doing seatwork.

Page 10: Race, social class, and student engagement in middle school English classrooms

Table 3Counts of question asking and answering among high and low SES students within classrooms, standard deviation in parentheses(weighted by # of questions per classroom observation)a

Questiontype:

Teacherinitiated

Studentinitiated

Authentic Authenticuptake

High LevelQuestion

ProceduralQ&A

SubstantiveQ&A

Low SES 11.92 (14.8) 1.44 (4.25) 6.53 (9.54) 2.76 (5.34) 3.32 (6.25) 6.27 (10.14) 6.98 (10.00)High SES 16.97 (16.6) 2.07 (4.73) 10.86 (11.87) 4.33 (6.34) 5.74 (8.36) 7.34 (10.45) 11.48 (12.31)

a Lowest and Highest quartile of students (N = 516).

S. Kelly / Social Science Research 37 (2008) 434–448 443

white students, have lower levels of participation in classroom discourse in predominantly black class-rooms. But, because black students are of course more likely to be found in predominantly black class-rooms, this negative class level effect means they experience fewer verbal exchanges with teachers thanwhite students.

5.1.2. The effects of initial achievement on engagementModels 2 and 4 in Table 2 indicate that initial achievement has a positive effect on both procedural and

substantive Participation. Students who enter class with strong reading and writing skills are more likely toparticipate in classroom discourse in English and language arts classrooms. A one standard deviation increasein reading and writing scores leads to a 1.35 factor increase in instances of substantive participation. The var-iance explained statistics for Models 1 and 2, and Models 3 and 4 in Table 2 reveal that the models includinginitial achievement explain larger proportions of the variance in participation than models that contain onlygender, SES, and race/ethnicity.

At the class level though, average achievement is unrelated to levels of student participation. Within class-rooms, higher achieving students are more likely to participate in classroom discourse, a relative effect. Butdifferences in absolute achievement levels across classrooms are unrelated to levels of procedural or substan-tive discourse. Examining the random effect of initial achievement within classrooms (across the set of class-rooms) in Table 2 reveals that the effect of initial achievement on participation is similar across classrooms.

I have conceived of participation in classroom discourse as an indicator of student engagement, with higherlevels of participation indicating higher levels of engagement. But perhaps higher achieving students partici-pate more because they are more likely to know the answers to questions? If this is the case than differentialparticipation may be pedagogically appropriate. A certain amount of differential participation could still ben-efit low achieving students, because they could learn from the responses of higher achieving students, espe-cially if students adopted a model of classroom discourse which valued the inputs of other students inclassroom instruction (Morine-Dershimer, 1985). If high achieving students participate more because theyknow the answers, then we might expect that the difference in participation between high and low achievingstudents would be greater on difficult questions with a high cognitive level. In fact, the difference is greater onlow-cognitive level questions to which many low achieving students probably do know the answers.9 Theobserved disparity in participation between high and low achieving students probably reflects a combinationof differences in knowledge and levels of engagement.

5.1.3. The effects of class composition

Models 5a and 6a in Table 4 examine the relationship between class composition and differential levels ofstudent engagement among students within classrooms using the reduced form models (without controls forinitial achievement). Does class composition affect the relative levels of engagement of black and lower socialclass students within classrooms? Is a low-SES student less likely to participate in class when they are sur-rounded by mostly low-SES peers? The conditional random effect variances indicate that there tends to bemore variation in these effects across classrooms when substantive participation is the dependent variable.In both cases though, the reliability of the slope effects is high enough that if there is a strong multi-level inter-

9 High achieving students (highest vs. lowest quartile) answered an average of 5.4 high level questions, and 13.3 low level questions, lowachieving students answered 4.7 and 10.6 high and low level questions, respectively.

Page 11: Race, social class, and student engagement in middle school English classrooms

Table 4Participation as a function of race and social class

Procedural participation Substantive participationa

5a 5b 6a 6bUn-centered Variables Un-centered Variables

Within-class

Male .28 (.049)*** .30 (.049)*** .21 (.046)*** .21 (.046)***

Hispanic �.11 (.087) �.091 (.086) .045 (.084) .043 (.084)American Indian .063 (.10) .078 (.10) �.20 (.11) �.19 (.11)Asian �.42 (.15)** �.41 (.15)** �.24 (.14) �.24 (.14)Other race �.21 (.27) �.27 (.27) .022 (.22) �.026 (.22)Black �.058 (.086) .018 (.10) .037 (.12) .075 (.12)

X class percent blackb �.13 (.39) �.57 (.39) .52 (.46) �.073 (.39)Random effect (SD) .45 .40 .54** .54**

SES scale .19 (.034)*** .15 (.037)*** .15 (.041)*** .16 (.040)***

X class SES scale �.16 (.098) �.16 (.11) .052 (.11) �.011 (.11)Random effect (SD) .15 .14 .23* .22*

Between-class

Percent black �.89 (.31)* �.45 (.40) �.44 (.23) �.55 (.33)SES scale �.83 (.28)** �.69 (.30)* .25 (.21) .082 (.23)

Poisson models, variable exposure, adjusted for over-dispersion. 1922 students in 117 classrooms.a Year of data, experimental cohort of classroom, and year by cohort interaction at the school level added to models of substantive

Participation. Coefficients not reported.b Multi-level interactions use grand-centered level 2 predictors. Those for Black X Class Percent Black use 70 of the 117 classrooms,

those that contain both white and black students.* P < .05.

** P < .01.*** P < .001.

444 S. Kelly / Social Science Research 37 (2008) 434–448

action between class composition and the within-class effect, it should be detected (.195 and.121 for the blackand SES coefficients in model 5, .296 and .272 for those coefficients in Model 6).10 The results indicate thatthere is no relationship between the composition of the class, and the within-class effect of being black, orof social class.

Recall from Table 2 that in predominantly black classrooms teachers and students engage in fewer verbalexchanges. Moreover, these results change only slightly when achievement at the class level is taken intoaccount because aggregate levels of achievement are not strong predictors of either form of classroom dis-course (Table 2). Models 5b and 6b reconsider the effects of class-level SES and the percentage of black stu-dents in a class, by using uncentered level-one variables. By comparing Models 5a and 6a to 5b and 6b, we canassess the contextual effect of social class and racial composition. In Models 5b and 6b, the entire effect of theclass level coefficient is contextual, meaning it affects all students within the class, regardless of race or SES.Focusing only on the class level coefficients that are significant, those for procedural participation, the com-parison across models indicates that about half of the class-level effect of racial composition is contextual,while most of the SES effect is contextual. Something about the context of predominantly black and low-SES classrooms leads to fewer verbal exchanges between teachers and students, which cannot be attributedsimply to the propensity of the individual students to ask and answer questions.

5.2. Student-reported measures of effort

Table 5 reveals some important similarities between the findings uncovered for participation in classroomdiscourse, and student-reported effort. First, as with participation in Q&A sessions, the majority of the var-

10 Reliabilities of .2–.3 would be considered prohibitively low for a scale, but for random coefficients in multi-level models, this level ofreliability is not unusual and is typically deemed acceptable (Bryk & Raudenbush, 1992, p. 110). Even though the signal in a randomcoefficient is often low, important predictors of these coefficients are often found in multi-level analyses.

Page 12: Race, social class, and student engagement in middle school English classrooms

Table 5Student reported effort as a function of race, social class, and standardized achievement score

7 8 9

Within-Classa

Male �.23 (.032)*** �.19 (.032)*** �.19 (.032)***Hispanic �.11 (.059) �.077 (.058) �.075 (.058)American Indian �.10 (.071) �.096 (.071) �.086 (.071)Asian .11 (.085) .11 (.084) .11 (.084)Black �.081 (.063) �.053 (.062) �.25 (.11)*

Class percent black .49 (.23)*SES scale .18 (.024)*** .16 (.024)*** .16 (.025)***

Class level SES .11 (.058)Achievement scale .033 (.01)*** .033 (.01)***

Random effect (SD) .030*** .029b

Variance explained (%) 5.7% 9.6% 11.5%

Between-class

Percent black .032 (.10) .026 (.10) .013 (.10)SES scale .43 (.074)*** .43 (.076)*** .43 (.076)***Achievement scale �.0038 (.013) �.005 (.013)Variance explained (%) 45.3% 42.9% 42.6%

1916 students in 116 classrooms.a All within-class variables besides gender are group-mean centered.b P-value = .057, because it is based on only 67 level two units, but the standard deviation (SD) is the same.

S. Kelly / Social Science Research 37 (2008) 434–448 445

iance in reports of student effort lies within classrooms (86.3%). This is striking when one considers that thestudent effort scale should reflect in part how much material was assigned by teachers in each class (writingand reading assignments, homework, and paper revisions), which might be expected to vary from class toclass. Second, the most important determinant of participation in Q&A, initial achievement, is also the mostimportant predictor of effort. As with participation in Q&A, there were few differences between effort reportedby black and Hispanic students, and that of whites, and social class was an important determinant of effort,independent of other variables.

Model 9 presents the multilevel interactions between classroom composition and the effects of student char-acteristics within the classrooms, similar to Models 5a and 6a for participation during Q&A. Recall that nodifferences in participation between blacks and whites or high and low SES students were found, regardless ofthe composition of the classroom. The same results hold for effort in Model 9. Indeed, in predominantly blackclassrooms, blacks report higher not lower levels of effort than they do in integrated classrooms.11 The resultsfrom Models 5a, 5b and 9 are inconsistent with social identity theories of peer sanctioning. It is unlikely thatdirect pressure from peers is causing black students in these predominantly white classrooms to exert (or atleast report exerting) less effort since very few of their peers are present in predominantly white classrooms.

Despite having several common determinants, the overall correlation between the effort scale and the obser-vational measures of procedural and substantive participation within classrooms is quite weak (.060, .046,respectively). There were a few notable differences in individual-level effects as well; boys are much more likelyto participate in classroom discourse, but less likely to report exerting effort on reading and writing tasks.Asians, who talk less in class, report similar levels of effort as white students. At the classroom level, perhapsthe most important finding in Table 5 is that the effects of initial achievement on student effort vary acrossclassrooms (random effect = .030) more so than do effects on participation in Q&A. In a few classrooms, rel-atively low-achieving students report putting forth equal amounts of effort as their higher-achieving peers. Per-haps this could be explained by differences in the nature of the assignments in different classrooms, or in howteachers evaluate students’ work.

11 The intercept for black students refers to black students in predominantly white classrooms, since the class level coefficient isuncentered in Model 4.

Page 13: Race, social class, and student engagement in middle school English classrooms

446 S. Kelly / Social Science Research 37 (2008) 434–448

6. Discussion

While there is no singular explanation for educational inequality, disparities in student engagement withinschools play an important role. Some students come to adopt the perspective that success in school is not forthem, and this is revealed both in negative sentiments towards school, and reduced effort and withdrawnbehavior. Educational researchers have provided compelling portraits of the process of withdraw being linkedwith students’ social identities, and peer group interactions. In this analysis, I searched for evidence of such aphenomenon by examining positive forms of student engagement within middle school English classrooms. Ireasoned that if anti-school norms are present among black or low-SES students, than students will feel morepressure to conform to those norms in predominantly black or low-SES classrooms. However, I found no evi-dence to support such a hypothesis.

There was evidence that low-SES students in particular were less engaged within classrooms, but these dif-ferences might be attributed to individual differences, such as dynamics within the family, as easily as collectiveaction among peers. For example, Lareau (2003) found differences in child-rearing practices between middleclass and working class/poor families where middle class children engaged in frequent verbal negotiations anddiscussions with adults. Middle class parents, she argues, cultivate their children’s verbal skills, teaching themto express their ideas, argue, and reach compromises in their interactions with adults. Middle class childrenmay be more practiced at voicing their ideas, and have a greater inclination to talk in class.

I also found that predominantly black classrooms had fewer instructional verbal exchanges, that there wassimply less talk between teachers and students. It is possible that this could be traced to social creativityamong students within those classrooms and sanctioning of peers who show interest in school. However,all students, including white students, talk less in predominantly black classrooms. Perhaps this finding reflectsthe approach taken by teachers in predominantly black classrooms, that classroom discourse is purposefullyrestrained. Tyson (2003) found that teachers in two predominantly black schools were preoccupied with thedeportment of their students. This stance could be traced to teachers’ beliefs that behavioral socialization iscrucially important to their students achieving success in higher education and the workplace. Whether thestruggle by teachers in predominantly black schools to maintain order in the classroom is a result of genuinelydisruptive behavior, or their own perceptions of behavior, the result may be the same, an instructionalapproach that focuses on maintaining classroom order and minimizing classroom disruptions. Teachers seek-ing to maintain an orderly environment are likely to actively avoid interactive whole-class forms of instruction(Metz, 1978).

Most importantly, the strongest predictor of engagement within classrooms is not race/ethnicity or socialclass, but a student’s level of initial achievement. Students who enter class with less advanced reading and writ-ing skills participate less often in classroom discourse and exert less effort on classroom tasks. These findingsare consistent with prior research on classroom discourse at the elementary school level (Morine-Dershimer,1985; Good et al., 1987) and a wide body of research on student engagement. Moreover, in these untrackedand regular-track middle school classrooms, only differences in achievement within classrooms affected studentengagement. This finding strongly suggests that the observed differences in engagement can be traced to thesocial setting of the classroom; to processes of social comparison and resultant self-perceptions of abilityand self-worth.

In this analysis, two very different measures of engagement were employed, detailed observations of studentparticipation, and student-reported measures of effort. In this sample of students these measures were nothighly correlated, and yet each outcome had common determinants, the most important of which was initialachievement.12 Engagement is a multidimensional concept, encompassing both an affective state as well asobservable behaviors such as participation in classroom discourse (Newmann, 1992). In the present study,the observational and student-report data both measured behavioral engagement. In addition to measurementerror associated with the student questionnaires, the low correlation between these two sets of measures maybe partly explained by the fact that some students who seldom participated in Q&A may have still been

12 The inequality in participation in classroom discourse seemed to be more intractable than differences in student effort. In the 117classrooms examined here, initial achievement was positively associated with participation in every classroom.

Page 14: Race, social class, and student engagement in middle school English classrooms

Table A1Descriptive statistics, 1922 students in 117 classrooms

Mean Standard deviation

Student level

Procedural engagement counts 3.09 4.92Substantive engagement counts 4.21 5.63Hispanic .09 .28American Indian .05 .23Asian .04 .19Other race .22 .41Black .81 .16Home resources .77 1.34Parental occupational education 3.9 1.10Parental education 1.20 .78SES scale 7.58 3.77

Between-class

Procedural engagement counts 3.25 2.95Substantive engagement counts 4.19 3.12Percent black .26 .34Home resources 1.16 .37Parental occupational education 7.42 2.25Parental education 3.09 4.92SES scale 4.21 5.63

S. Kelly / Social Science Research 37 (2008) 434–448 447

engaged in the lesson, and these same students may have readily completed classroom tasks and homeworkassignments. Observational measures of student engagement are an excellent way to uncover positive instancesof engagement, or highly-withdrawn and disruptive behavior. Questionnaires can be very useful in identifyingthe sources of student engagement as well, particularly if they include affective measures of interest, concen-tration, and enjoyment during classroom instruction (Shernoff et al., 2003). Unfortunately, such affective mea-sures are not available in the Partnership data.

Social scientists and educators should continue to be concerned about the relationship between initialachievement and student engagement. To the extent that race, social class, and initial achievement are corre-lated, breaking the relationship between achievement and engagement will help traditionally disenfranchisedstudents experience more effective instruction.

Appendix A

Table A1

References

Ainsworth-Darnell, James W., Downey, Douglas B., 1998. Assessing racial/ethnic differences in school performance. AmericanSociological Review 63, 536–553.

Applebee, Arthur N., Langer, Judith A., Nystrand, Martin, Adam, Gamoran, 2003. Testing a framework for effective English instruction:a study of relationships between achievement gain and instructional approaches in middle and high school classrooms. AmericanEducational Research Journal 40 (3), 685–730.

Bandura, Albert, 1977. Self-efficacy: toward a unifying theory of behavioral change. Psychological Review 84, 191–215.Berends, Mark., 1995. Educational stratification and students’ social bonding to school. British Journal of Sociology of Education 16,

327–351.Bryk, Anthony S., Stephan W. Raudenbush. 1992 [2002]. Hierarchical Linear Models. Newbury Park: Sage Publications.Carter, Prudence L., 2005. Keepin It Real: School Success Beyond Black and White. Oxford University Press.Cohen, Elizabeth G., Lotan, Rachel A., 1997. Working for Equity in Heterogenous Classrooms. Teacher’s College Press, Columbia.Cook, Phillip J., Ludwig, Jens, 1997. Weighing the ‘‘Burden of Acting White:’’ Are there race differences in attitudes toward education?

Journal of Policy Analysis and Management.

Page 15: Race, social class, and student engagement in middle school English classrooms

448 S. Kelly / Social Science Research 37 (2008) 434–448

Covington, Martin, Berry, Richard G., 1976. Self-Worth and School Learning. Holt, Rinehart and Winston, New York.Datnow, Amanda, Cooper, Robert, 1997. Peer networks of African American students in independent schools: affirming academic success

and racial identity. The Journal of Negro Education 66, 56–72.Eckert, Penelope, 1989. Jocks and Burnouts: Social Categories and Identity in the High School. Teachers College Press, New York.Finn, Jeremy D., 1989. Withdrawing from school. Review of Educational Research 59, 117–142.Finn, Jeremy D., Pannaozzo, Gina M., Voelkl, Kristin E., 1995. Disruptive and inattentive-withdrawn behavior and achievement among

fourth graders. Elementary School Journal 95, 421–434.Fordham, S., Ogbu, John, 1986. ‘‘Black Students’ School Success’’: soping with the ‘‘Burden of Acting White’’. The Urban Review 18,

176–206.Gamoran, Adam, Mare, Robert D., 1989. Secondary school tracking and educational equality: compensation, reinforcement, or

neutrality. American Journal of Sociology 94, 1146–1183.Gibson, Margaret A., 1987. The school performance of immigrant minorities: a comparative view. Anthropology & Education Quarterly

18, 262–275.Goffman, Erving, 1963. Stigma: Notes on the Management of Spoiled Identity. Prentice Hall, Englewood Cliffs, N.J.Good, Thomas L., Slavings, Ricky L., Havel, Kathleen H., Hugh, Emerson, 1987. Student passivity: a study of question asking in K-12

classrooms. Sociology of Education 60, 181–199.Hauser, Robert M., Warren, John R., 1997. Socioeconomic indexes for occupations: a review, update, and critique. Sociological

Methodology 27, 177–298.Hechter, Michael, 1987. Principles of Group Solidarity. University of California Press, Berkeley.Horvat, Erin M., Lewis, Kristine S., 2003. Reassessing the ‘‘Burden of ‘Acting White’’’: the importance of peer groups in managing

academic success. Sociology of Education 76, 265–280.Langer, J.A., Applebee, A.N., Nystrand, M. 2005. Final report. National Research Center on English Learning and Achievment. SUNY-

Albany.Lareau, Annette, 2003. Unequal Childhoods: Class, Race, and Family Life. University of California Press, Berkeley.MacLeod, Jay, 1987. Ain’t No Makin’ It: Leveled Aspirations in a Low-Income Neighborhood. Westview Press, Boulder, CO.McFarland, Daniel A., 2001. Student resistance: how the formal and informal organization of classrooms facilitates everyday forms of

student defiance. American Journal of Sociology 107, 612–678.McLaren, Peter, 1986. Schooling as a Ritual Performance. Routledge and Keegan Paul, London.Metz, Mary H., 1978. Classrooms and Corridors. University of California Press, Berkeley.Mickelson, Rosyln A., 1990. The attitude-achievement paradox among black adolescents. Sociology of Education 63, 44–61.Morgan, Stephan L., Mehta, Jal D., 2004. Beyond the laboratory: evaluating the survey evidence for the disidentification explanation of

black-white differences in achievement. Sociology of Education 77, 82–101.Morine-Dershimer, Greta, 1985. Talking, Listening and Learning in Elementary Classrooms. Longman, New York.Newmann, Fred M., 1992. Student Engagement and Achievement in American Secondary Schools. Teachers College Press, Columbia.Nystrand, Martin, 1997. Opening dialogue: Understanding the Dynamics of Language and Learning in the English Classroom. Teachers

College Press, Columbia.Nystrand, Martin, Gamoran, Adam, 1991. Instructional discourse, student engagement, and literature achievement. Research in the

Teaching of English 25, 261–290.Ogbu, John U., 1990. Minority education in comparative perspective. The Journal of Negro Education 59, 45–57.Patchen, Martin, 1982. Black-White Contact in Schools: Its Social and Academic Effects. Purdue University Press, West Lafayette, IN.Roscigno, Vincent J., 1998. Race and the reproduction of educational disadvantage. Social Forces 76, 1033–1061.Rosenbaum, James E., 2001. Beyond College for All: Career Paths for the Forgotten Half. Russell Sage, New York.Roeser, R.W., Strobel, K.R., Quihuis, G., 2002. Studying early adolescents’ academic motivation, social-emotional functioning, and

engagement functioning, and engagement in learning: variable and person-centered approaches. Anxiety, Stress, and Coping 15, 345–368.Roeser, R.W., Eccles, J.S., Sameroff, A.J., 2000. School as a context of early adolescents’ academic and social-emotional development: a

summary of research findings. The Elementary School Journal 100, 443–471.Ruble, Diane N., Frey, Karen S., 1987. Social comparison and self-evaluation in the classroom: developmental changes in knowledge and

function. In: Masters, J.C., Smith, W.P. (Eds.), Social Comparison, Social Justice, and Relative Deprivation. Lawrence Erlbaum, NewJersey.

Shernoff, D.J., Csikszentmilhalyi, M., Schneider, B., Shernoff, E.S., 2003. Student engagement in high school classrooms from theperspective of flow theory. School Psychology Quarterly 18, 158–176.

Smith, William P., Davidson, Emily S., Ann-Claire, France-Kaatrude, 1987. Social comparison and achievement evaluation in children.In: Masters, J.C., Smith, W.P. (Eds.), Social Comparison, Social Justice, and Relative Deprivation. Lawrence Erlbaum, New Jersey.

Tajfel, Henri., John C. Turner. 1979 [1986]. An Integrative Theory of Intergroup Conflict. In: Worchel, S., Austin, W.G. (Eds.), The socialpsychology of intergroup relations, 2nd ed. Nelson-Hall, Chicago, pp. 33–47, pp. 7–24.

Tyson, K., Darity, W., Castellino, D., 2005. Black adolescents and the dilemmas of high achievement. American Sociological Review 70,582–605.

Tyson, Karolyn, 2003. Notes from the back of the room: problems and paradoxes in the schooling of young black students. Sociology ofEducation 76, 326–343.

Valeski, T.N., Stipek, D.J., 2001. Young children’s feelings about school. Child Development 72, 1198–1213.Voelkl, K.E., 1997. Identification with school. American Journal of Education 105, 204–319.Willis, Paul. 1977 [1981]. Learning to Labor: How Working Class Kids get Working Class Jobs. New York: Columbia University Press.