The Relative Importance of Home and School in the Development of Literacy Skills for Middle-Grade Students

by Valerie E. Lee, Robert G. Croninger
The Relative Importance of Home and School in the Development of Literacy Skills for Middle-Grade Students
Valerie E. Lee, Robert G. Croninger
American Journal of Education
Start Page: 
End Page: 
Select license: 
Select License

The Relative Importance of Home and School in the Development of Literacy Skills for Middle-Grade Students

VALERIE E. LEE and ROBERT G. CRONINGER University of Michigan

Variations in the home environments of poor and middle-income chil- dren affect their literacy development, which leads to substantial differ- ences in reading ability and behavior. Schools can mediate influences from home through the conditions that they foster and the instructional policies and procedures they promote. The result of schools' efforts may either ameliorate or magnify the inequities in reading development related to family economic conditions. This study tests these contentions in middle-grade schools by using a nationally representative sample of poor and middle-class eighth graders from the National Education Longitudinal Study of 1988 (NELS:88). Home and school effects on our measure of literacy develop~nent-a standardized test of reading com- prehension-are explored with ~nultilevel methods (hierarchical linear modeling). While homes exert an important influence on this outcome, findings focusing on schools and classrooms are emphasized. The study also highlights school conditions and policies that foster social equity in the literacy development of young adolescents. Implications of current school reform efforts are discussed.


Literacy is perhaps our most basic expectation for students during their elementary and secondary school years. Unfortunately, family poverty is related to children's school performance (Allington 1990; Natriello et al. 1990), particularly in reading and literacy. Poor children begin to fall behind their more affluent peers around the fourth grade. This deficit increases through the eighth grade, as students make the

American Journal of Education 102 (May 1994)
O 1994 by The University of Chicago. All rights reserved.
01 95-674419410203-0002$0 1 .00

286 American Journal of Education transition from "learning to read" to "reading to learn" (Chall et al. 1990; Snow et al. 1991). Although students from all social backgrounds fail to develop the requisite literacy slulls in the upper grades, this failure is more pronounced for students from economically disadvan- taged families. Indeed, the average reading proficiency of disadvan- taged eleventh graders is comparable to that of the average seventh grader (Applebee et al. 1988).

Common explanations for why poor children fall behind focus on either their home environments or their school experiences. Although studies have argued convincingly for the importance of each sphere in the development of literacy (e.g., see Scott-Jones 1984; Knapp and Needels 1990), they have seldom examined how different home envi- ronments and school experiences interact in influencing literacy out- comes. A notable exception is a longitudinal study conducted by Chall and Snow of the literacy development of elementary school children from low-income families (Chall et al. 1990; Snow et al. 1991). Al- though the study's external validity is limited (the sample included only 30 families from a single metropolitan area), their findings indi- cate that poor children's literacy failures result from an interplay be- tween home and school experiences. Moreover, school and home in- fluences differ for different literacy outcomes.

Most important in the context of the present study are the findings of Chall, Snow, and their colleagues (Chall et al. 1990; Snow et al. 1991) that poor children are especially dependent on school-related opportunities to develop literacy skills, particularly when home sup- ports are weak or ineffectual. We expand on these researchers' im- portant findings about the literacy development of children living in poverty with the use of a large and nationally representative sample of poor and middle-income eighth graders. We explore home and

VALERIEE. LEE is associate professor of education at the University of Michigan, where she teaches courses in quantitative methods and the sociology of education. Her current research focuses on school restructuring, school choice, and equity in educational outcomes by race and ethnicity, social class, and gender. ROBERT G. CRONINGER

is a doctoral student in educational studies at the University of Michigan, and an associate director of the Programs for Educational Opportu- nity, a Title IV desegregation assistance center. He is currently study- ing organizational practices and policies that shape the educational experiences of protected and disadvantaged populations.

May 1994 287

school influences on young adolescents' performance on an important literacy outcome: reading comprehension.


Forms of Literacy

Literacy is more than reading; rather, it involves a wide range of cultural practices and activities. Although the competencies encom- passed by these practices are related, they are not identical, especially at more proficient levels of performance. We distinguish between liter- acy as a general set of skills, useful in all contexts and all literacy- related activities, and literacy as a more refined set of competencies connected to the context in which people engage print. The conceptu- alization of literacy in this study emphasizes the latter. The broader view implies that schools represent only one source of literacy develop- ment. The form of literacy in question determines, in large part, the sources that may provide the most powerful opportunities for learning.

The form of literacy on which we concentrate is what Resnick (1991) calls informational literacy (using print to acquire and convey new knowledge).' This form is most clearly associated with schools, al- though the extent to which schools and classrooms are structured to meaningfully promote its development has been questioned (Dole et al. 1991 ; Newmann 1991 ; Hull 1989). Reading comprehension is an important dimension of this form of literacy, for it determines the extent to which print is accessible to students. At about the fourth grade, the demands on reading comprehension intensify as students are assigned increasingly complex materials with varying formats (Chall et al. 1990). These demands increase through the middle grades and into high school, as constructing meaning and monitoring under- standing become the qualities that distinguish good readers from poor readers (Dole et al. 1991).~

Home Supports for Literacy Development

There is considerable evidence that parents' literacy practices, support for learning at home, and involvement in school activities affect literacy development, even after students' socioeconomic status and abilities are taken into account (Scott-Jones 1984; Epstein 1991). Snow et al. (1991) found that the home literacy environment and the parents'

288 American Journal of Education involvement at school were significantly related to the reading ability of low-income elementary school students. Parents who read fre- quently, who sought out literacy experiences for their children, who attended school meetings and events, and who volunteered at school had children who gained more in reading comprehension than parents who did not engage in these activities. It is clear that the home can be an important source of literacy development for all children, re- gardless of economic conditions.

Parents vary, however, in their ability to provide support. Although virtually all parents want to help their children do well in school, not all parents are capable of doing so (Epstein 1986). Minority group parents, for example, have exceptionally high educational aspirations for their children, but they often lack the material resources required to support those aspirations (Stevenson et al. 1990). Furthermore, minority group parents who have historically experienced discrimina- tion (especially blacks, Hispanics, and Native Americans) find it more difficult to convince their children that extended effort will be re- warded with social and economic success (Ogbu 1991).

Moreover, parents of poor children have less direct experience with schools than the parents of advantaged children. Many poor parents failed to complete high school themselves or, if they did, remember school as a difficult and trying experience (Natriello et al. 1990). In the early grades, differential educational background is less critical, for school-related tasks are relatively simple and school structure more personal. As children advance in grade, school tasks increase in com- plexity, sometimes exceeding the ability of parents to perform them. More formal education renders parents better able to assist their chil- dren with homework, to advocate for and acquire special services when necessary, and to help their children consider alternative strategies and solutions (Epstein 1986; Stevenson and Baker 1987; Lareau 1989). Because mothers are more likely to manage children's education than fathers, the maternal role is central in determining the quantity and quality of home supports that children experience (Baker and Steven- son 1986; Lareau 1989).

School Supports for Literacy Development

Besides deficient home environments, a primary explanation for the underdevelopment of literacy skills among poor children has long been the belief that school resources and facilities are distributed unequally. Although earlier studies, such as Coleman et al.'s (1966) Equality of Educational Opportunity, found minimal school effects, more recent

May 1994 289 studies, which use more appropriate methodologies, have found strong school influences on student outcomes, especially on a~hievement.~ We consider three categories of school attributes, in order from those least responsive to those most responsive to school reforms and initia- tives: school composition, school conditions, and instructional policies and practices.

School composition.-Poor students frequently attend school with other poor students (Orland 1990). Often, these schools have high minority enrollments, particularly if they are located in urban areas (Comer 1988; Wilson 1987). Large proportions of poor and minority- group children create particular difficulties for schools (Barr and Dreeben 1983; Coleman et al. 1966; Cusick 1983). Students in these schools are aware of the social and educational inequalities that they face, which often undermines their motivation and encourages the formation of peer cultures in opposition to academic goals (Kozol 1991). High concentrations of disadvantaged students also pose unique problems in the organization of classroom instruction, as these students often require more assistance and make it difficult for teachers to maintain a good instructional pace (Barr and Dreeben 1983; Cus- ick 1983).

School sector and size can mediate the relationship between family status and achievement. Lee and Bryk (1989), for example, found that the relationship between social class and mathematics achievement was different in public and Catholic high schools. Average social class plays a more powerful role in public schools than Catholic schools, and disadvantaged students tend to achieve at higher levels in Catholic schools than public schools. Stratification in achievement by students' social class, however, is more likely to occur in large than small schools, regardless of sector (Lee and Bryk 1989; Lee and Smith 1993). Eccles et al. (1991) argue that large middle-school structures present young adolescents with difficult developmental challenges. They found these environments to have an adverse effect on student discipline, engage- ment, achievement, and perceptions of self, especially for students who have experienced earlier school difficulties.

School conditions.-The composition of schools is often a proxy for school conditions that are more directly associated with literacy devel- opment. We suggested above that high concentrations of disadvan- taged students and large school size, for example, often create environ- ments that disengage many students, which distracts them from learning. Nonetheless, other school conditions may have more direct effects on literacy development than school composition. Comer (1980, 1988) argues that poor and minority children do better in schools that explicitly attempt to bridge the social and cultural gap

290 American Journal of Education between home and school. Other studies have found benefits for stu- dents by encouraging the involvement of poor parents in school and by supporting their involvement in learning at home (see, e.g., Ascher 1988; Epstein 1991). The most dramatic effects may occur in those schools where strong family-school ties lead to the formation of what Coleman (1987) calls "functional communities." Parents and teachers are more likely to know each other in functional communities, to develop positive relations, and to work together toward goals that they all share. Their increased interaction creates "social capital" (Coleman 1988), which permits families and schools to use one another more effectively in monitoring student behavior and shaping student development.

The school effectiveness literature suggests that orderly school envi- ronments and positive student-teacher relationships also have positive effects on learning, particularly in urban schools with high concentra- tions of disadvantaged students (Hallinger and Murphy 1986). An orderly environment allows schools to maintain a focus on learning, to clearly communicate expectations for student behavior, and to main- tain a disciplinary climate that facilitates learning. Positive student- teacher relationships mean that students may experience their schools as caring communities and see their teachers as personally interested in their well-being and progress. The quality of student-teacher rela- tionships can be especially important to young adolescents, who are sometimes traumatized by the impersonal and bureaucratic environ- ments that characterize traditional middle-school structures, particu- larly when compared to elementary schools (Eccles et al. 1991). Cer- tainly a potential exists for conflict among these conditions-schools can be either "too orderly" or "too personal and student-directed" for students to engage in learning. The critical developmental period of adolescence requires a delicate organizational balance in schools be- tween social integration and individuality (Newmann 1981; Eccles et al. 1991). Middle-grade schools must consider both of these needs, while also promoting the primacy of academic goals.

Another set of extremely difficult conditions for schools includes excessive teacher absenteeism, student absenteeism, and concerns about personal safety. These conditions characterize a school's norma- tive environment, which has been shown to affect learning. In schools with safe and orderly environments, as well as low levels of staff prob- lems (which includes teacher absenteeism), the achievement of disad- vantaged students is high (Lee and Bryk 1989). High levels of student absenteeism certainly disrupt instruction, forcing teachers to juggle lesson plans when a significant proportion of students are absent (Barr and Dreeben 1983). Although each of these conditions is dis-

May 1994 291 tinct, they are similar in that each takes time and energy away from the school's technical core and fundamental task-instruction (Bidwell 1965).

School policies and practices.-Schools' policies and practices surrounding instruction are likely to affect literacy development more directly than their compositions or conditions. Although school com- position and condition establish the context in which instruction oc- curs, it is instruction itself that determines what children learn (Barr and Dreeben 1983). Not surprisingly, Chall et al. (1990) found instruc- tional practices and classroom provisions for literacy enrichment to be significantly related to poor children's gains in reading comprehen- sion. In classrooms where comprehension was explicitly taught, where time was set aside for students to read, where students were taken to visit the library, and where a wide range of reading materials was available, children from low-income families made the greatest gains in reading comprehension. Less important but still exerting a small positive effect on comprehension gains was a measure of classroom emphasis on higher-level processes (e.g., making inferences or devel- oping metacognitive strategies). The authors argue that these weak findings may have resulted from the fact that the use of metacognitive strategies only begins to appear in the seventh and eighth grades (Haller et al. 1988), which was at the upper age limit of their subjects (who were followed primarily in the elementary grades).

A number of studies have reached a similar conclusion: poor chil- dren become more competent in literacy when reading for meaning is emphasized, when writing tasks emphasize authentic acts of commu- nication, when these children are exposed to a wide variety of texts (especially those that connect with students' backgrounds and experi- ences), and when there is more than trivial emphasis on higher-order processes and the thoughtful engagement of students (Newmann 1991; Knapp and Needels 1990). These findings fly in the face of more conventional beliefs that poor students require lower-level cur- riculum and the careful sequencing of instruction, which breaks com- petencies into small, discrete bundles of basic skills that can be mas- tered separately through drill and practice (Knapp and Turnbull 1990). The emphasis is, rather, on engaging students in complex and challenging tasks, on helping students regulate their own learning, and on stimulating higher levels of performance through interactions with peers, computers, and teachers (Hull 1989; Dole et al. 1991). Although most of the existing evidence for the value of higher-order instruction for learning comes from ethnographic inquiries, small- scale studies, or evaluations of unique experiments in curriculum and instructional design, a few studies with greater external validity have

292 American Journal of Education reached similar conclusions, even after controlling for the socioeco- nomic status and ability of students (e.g., see Gamoran 1991).

Other policies and practices that help to shape school supports for literacy development focus on the school's academic organization and the differentiation of instruction by ability. Student course taking and tracking, for example, are the most powerful predictors of academic achievement, far stronger than the effects of either personal back- ground or students' attitudes and behaviors (for a review of this litera- ture, see Lee et al. [1993]). In schools that organize curriculum by ability or previous academic performance, disadvantaged students are typically found in lower-level classes. These classes cover less content, emphasize drill and practice, and often require teachers to spend more time on managing student behaviors (Knapp and Turnbull 1990; Oakes 1985). Although there is evidence that low-achieving students can benefit from flexible, temporary groupings that provide them with specialized attention and support, more permanent groupings that separate these students from higher achieving students severely re- strict their opportunities to develop literacy skills (Allington 1990; Knapp and Needels 1990; Slavin 1987). We anticipate, therefore, that poor children will do less well in schools that differentiate instruction by student ability.

Social and professional relations among teachers around instruction are important elements in a school's organization. Greater teacher collaboration and teacher teaming allow for extended interactions with students across subject areas and grades, which is hypothesized to increase student engagement, as well as teacher commitment and com- petence (Bryk et al. 1993; Newmann 1993). Team teaching has also been shown to be positively related to achievement (Lee and Smith 1993). Informational literacy, as described here, goes well beyond a simple concern about declarative knowledge to include the processes through which students acquire meaning, synthesize ideas, convey knowledge to others, and monitor their own understanding (Hull 1989; Dole et al. 1991). It is reasonable to conclude that inducing competency of this type in students requires considerable involvement and support from teachers, which is more likely to occur in schools where students and teachers have more frequent and more "authentic" interactions (Newmann 1991; Resnick 1991).

Research Questions

The purpose of this article is to compare the home and school supports for literacy for children from families living under different economic

May 1994 293 conditions. Given the complexity of this task, we pose a series of pro- gressively more complex research questions.

Question 1.-How do the home and school conditions of poor and middle-class young adolescents compare! How do the literacy develop- ments of these groups compare, as defined by reading for comprehension!

Question 2.-What effect do the differential levels of poor and mid- dle-class children's home support for literacy have on reducing differ- ences in literacy between these groups! Do home supports have different effects on literacy development for these two groups!

Question 3.-Which elements of middle-grade schools' supports for literacy (especially school conditions, policies, and practices) are related to poor and middle-class children's reading achievement, once differ- ences in home conditions are accounted for!

Question 4.-Which school supports for literacy are best able to explain the difference in reading achievement between poor and middle- class young adolescents! What, in other words, are the elements that foster social equity in literacy development!


Sample and Data

The sample for this study was drawn from the base year of the National Education Longttudinal Study of 1988 (NELS:88), a general-purpose study of the educational status and progress of about 25,000 eighth- grade students in 1,035 American middle-grade schools, sponsored by the National Center for Education Statistics (NCES; Ingels et al. 1989). The NELS base-year data follow a nested or stratified data structure, whereby schools were first sampled, and then a fixed number of stu- dents was sampled within each school. Data on students were collected from several sources: (1) a broad-based survey from students; (2) achievement tests in mathematics, science, reading, and social studies;

(3) a parent survey (usually of the mother); and (4) data from two of their teachers (either English or social studies, and either mathematics or science). Data describing schools were collected from principals, either through the normal NELS data collection or in a supplementary data collection in 1989.4 Students, teachers, and parents also describe their schools, and their descriptions may be used in aggregated form.

The study employs a subsample of these data on students and their schools, which was selected with the use of several data filters: (1) only

294 American Journal of Education students with data from their English teachers were included (this filter eliminated a random half of the NELS middle-grade schools and the sampled students in them); (2) only poor and middle-class students, defined as a family of four with an income below $48,250 (equivalent to the bottom three quartiles of a variable that is defined as the ratio of family income to 1987 federal poverty standards, adjusted for family size); (3) only schools with at least 10 NELS-sampled students; and

(4) students with data from all NELS data sources (students, schools, parents, and teachers). This resulted in a close to random subsample of 6,099 students in 377 schools, which averaged 16.2 students per school. Because the original NELS design oversampled certain types of schools (private schools and schools with high concentrations of Hispanic and Asian students), the NELS design weights (for schools) are employed for all analyses.


Dependent memure.-The study focuses on literacy development de- fined by a 21-item standardized test of reading achievement, with its scale equated by item response theory (IRT). This test "consists of five short passages followed by comprehension and interpretation ques- tions" (Ingels et al. 1989, p. 13), which students had 21 minutes to complete. This test has a mean of 10.03 and an SD of 5.84 in our sample. The psychometric properties of all variables used in this study, details of their construction, and original NELS component items from which they were constructed are provided in the Appendix.

Independent measures: deJining students. -To address the complexity of our research questions, the study includes several sets of indepen- dent measures categorized along different dimensions. One dimension contrasts measures that characterize students and their families (i.e., individuals) with measures describing their schools (i.e., groups). Three types of independent variables tap characteristics and behaviors of students and their families. The major analytic contrast for the study compares students defined by their family's economic condi- tion-in particular, whether they live in poor or middle-class condi- tions. We define "poor" as the lowest quartile on a continuous variable measuring the ratio of family income to 1987 poverty standards ad- justed for family size, and "middle class" as the middle two quartiles on this ~ariable.~

The second type of student-level independent vari- ables are demographic-mainly used as statistical controls. These in- clude academic background6 and measures of two sorts of minority status: raciallethnic status (black and Hispanic students are combined,

May 1994 295 as, unfortunately, this is substantially related to family economic condi- tions) and language minority status (given the study's focus on lan- guage and literacy skills).

A third set of measures defines the study's first major construct-home supports for literacy. This set includes parental status sup- ports (parents' educational level and mother's expectations for the student's educational attainment) and variables that define the educa- tional structure provided by families for their children. Here, several measures are available: use of the public library by the child andlor parent, literacy resources available in the home, and a factor describing the frequency with which students discuss school-related matters with their families.

Independent measures: dejning schools. -Our second major construct, school support for literacy, is operationalized by measures defining schools. These are drawn from several sources: the two data files from principals, as well as aggregates of variables defining schools and the learning environment therein from students, parents, and English teachers. A special strength of NELS is the data from teachers, in which they define actual English classes and how their students per- form there. Data from teachers linked to students, never before avail- able from NCES, are ideal to define the schools' learning environment for literacy. Our school variables, many of which are factors or compos- ites containing several submeasures, are also divided into sets that we have ordered to suggest the degree to which they may be changed by schools (from "difficult to change" to "explicit policies"). The first set describes school composition and structure (the proportion of poor and minority children, whether the school is located in an urban area and1 or a school sector [public, Catholic, or other private], whether the school is a stand-alone middle school or includes elementary and!or secondary grades, and the enrollment in the eighth grade).

A second set characterizes school conditions-variables that describe the school's environment and organization beyond "who goes there." These include a measure of home support for learning, parental satis- faction with the school, two measures of absenteeism (of English teach- ers and of students), the orderliness of a school's environment, positive teacher-student relations, a measure of social capital (measuring how many of the parents of their children's school friends are known by students' parents), and a measure of school safety. A third group of variables defines school policies and practices. These measures describe the frequency of authentic instruction in eighth-grade English classes (writing reports and compositions, use of literature, editing and rewrit- ing, oral presentations), the number of books used in eighth-grade English classes in addition to the textbook, cooperation and coordina-

296 American Journal of Education tion among teachers, and the degree to which the structure of classes is not grouped by ability. More detail on the construction of all vari- ables is provided in the Appendix.

Analytic Approach

Descriptive analyses. -Given the complexity of the data structure and the research questions, our approach to data analysis proceeds from descriptive through bivariate to multilevel. We begin by providing means for all variables considered in this study for poor and middle- class eighth graders-our target and comparison groups. We have tested (with t-tests) mean differences between the groups.

Bivariate relationships. -To build the final multivariate models, we investigated bivariate relationships (i.e., correlations) among the many student- and school-level measures described above. We provide sepa- rate correlations among the variables employed in our final analytic models for the variables describing individuals and those describing schools. To limit the amount of numerical information, we do not provide correlations among variables considered but not included in final models.

Multilevel analysis of home support.-A major purpose of the study is to contrast how supports for literacy in the homes of children from poor and middle-class families are related to reading achievement. Because it is probable that even these home supports (i.e., variables that characterize the students and their families within each school) vary systematically by the schools poor and middle-class children at- tend, these analyses are conducted with the use of a methodology identified as appropriate for what is known as "school effects research." The method, which is detailed below, is multilevel in nature. All home support measures in our multilevel models have been identified as important in research on poor children's literacy development.

Multilevel analysis ofschool support. -Our most important research ques- tions require investigations about how differences between schools- in their conditions, practices, and policies-influence the literacy de- velopment of their students. Because both our data and research ques- tions are characterized by a nested structure, it is important to employ a methodology that takes this structure into account. Thus, the major analytic method used for investigating both home and school supports is hierarchical linear modeling (HLM). Both the statistical theory and the approach underlying this methodology are explained in detail else- where, as is its superiority over ordinary least squares (OLS) regression as a tool to explore relationships at several levels-for individuals and

May 1994 297 groups and (equally important) between levels (Bryk and Raudenbush 1992; Lee and Bryk 1989). Before discussing empirical results, we present a brief exposition of the statistical procedure aimed at facilitat- ing the understanding of readers unfamiliar with the HLM methodology.

Overview of the Hierarchical Linear Model

Two types of outcomes. -To investigate school effects on reading com- prehension, we may characterize them either as mean differences between schools or as distributive effects within schools. Examples of distributive effects are those used by Lee and Bryk (1989) in their study of the social distribution of mathematics achievement, where they were interested not only in school mean math achievement but also in its equitable distribution within schools. "Social distribution" outcomes, which may be typified as relationships within schools, were characterized in that study as (a) the relationship between social class and achievement or (b) the minority achievement gap (i.e., the differ- ence in achievement level between minority and white students in each school). In this study, we are interested not only in school mean read- ing achievement, but also in the "poverty gap" in this outcome-that is, the mean difference in reading achievement between poor and middle-class students in each school. Given our hypotheses about dif- ferences in the effects of school variables for poor and middle-class children, examining causes of the poverty gap allows us to empirically test those hypotheses.

The generalform.-In its simplest form, HLM consists of two equa- tions, a within- and a between-unit model. Hierarchical linear modeling may be conceptualized as a set of single, small OLS regressions among students in each of a number of schools and as a larger regression run between schools. Coefficients estimated within each school become outcome variables in between-school equations. Our within-unit model represents reading achievement for student i in school j (Y,])as a function of various student characteristics, Xyk,and random error, e,:

The P,, regression coefficients (betas) are structural relations within school j that indicate how achievement is distributed with regard to measured student and home characteristics (the X,, variables). The

coefficients capture the social distribution of achievement in each school. A distinctive feature of HLhl is that these structural relations-

298 American Journal of Education the intercept as well as the slopes-are presumed to vary across units. Therefore, the between-unit model represents the variability in each Pjk parameter as a function of school-level variables, Wpl:

structural    effects of school-level    unique random
relations    characteristics on within-    effect associated
in unit j    school relationships    with school j
The ykp (gamma) coefficients represent the effects of school-level vari- ables (Wh) on the structural relations within unit j. In this case, the W variables measure aspects of school composition and structure, school conditions, and school policies and practices. The gammas are the effects of these school-level characteristics on the social distribution of achievement within schools.

Our model.-One of our within-school models examines reading comprehension for student i in school j as a function of poverty status, minority and language minority status, and academic background:

Reading Achievement+ = Pq + PllPoverty Status + P,jMinority Status

+ P3jLanguage Minority + P4jAcademic Background + eg.

Here, we characterize the social distribution of achievement in each school in terms of two parameters-an intercept (Pq) and a regression slope (PI,). Minority status, language minority status, and academic background are included as statistical controls only.7 Academic back- ground is a continuous variable centered around the mean for all schools. Poverty status, minority status, and language minority are dummy variables representing poverty, minority group membership, and home language background. As a result of this choice of metrics, the beta parameters of interest may be interpreted as follows:

pq = Mean reading achievement for middle-class students in

schoolj, which we refer to as the base achievement level

in school j. p,j = The mean difference between the achievement of poor and middle-class students in school j. This is referred to as the poverty gap.

Under this model, a school that is effective in equalizing achievement would be characterized as simultaneously having a high mean achieve- ment level (Pq) and a small poverty gap (P,~).~

These two distributive parameters are hypothesized to vary across schools as a function of

May 1994 299 school-level differences in composition and structure, conditions, and policies and practices. Separate between-school equations are thus posed for each beta coefficient. School characteristics promoting an effective and equitable distribution of achievement would demonstrate the following pattern of statistical associations in the between-school model: (a) a positive relationship to average reading achievement, and

(b) a positive effect on the poverty gap (as our coding scheme [poor = 1,middle class = 0] results in a poverty gap that is typically negative). That is, ideal school factors would simultaneously act to raise achieve- ment and to reduce the differences in achievement between poor and middle-class students in a school. Our HLM models involve a search for such variables.

Model building. -Given the large number of measures describing schools considered in this study, we developed a model-building strat- egy for selecting variables to include in our final HLM models. We considered school-level variables by the sets shown in table 1 (school composition, school conditions, school policies and practices). We first began with the full set of composition variables, removed variables one by one whose relationships produced t-statistics of 1.5 or less, and reestimated the model after each change. After an appropriate school composition model was developed, we then considered the set of school condition variables by using the same elimination criterion and reest- imation procedure. Finally, we considered the variables measuring policies and practices. As more variables were added to the model, others fell below our statistical significance criterion and were dropped. We followed this incremental strategy to develop the HLM school-level model in this study.


Descriptive Differences between Poor and Middle-Cluss Children

Variability among children and families. -The family, home, and school environments of young adolescents from poor and middle-class fami- lies are quite different, as group means in table 1 show. With few exceptions, means differ at probability levels less than .OOl.' The ana- lytic definition of our "poverty" category has led to our sample's being composed of one-third poor children (n = 2,037) and two-thirds mid- dle-class children (n = 4,062). Poor children score .5 SD below their middle-class counterparts in the test of reading comprehension, which is a large difference.'' In addition to family income (the basis of group

300 American Journal of Education

Croup Means for Poor and Middle-Class Eighth Graders on All Model Variables

Poor Middle-class Children Children (n = 2,037)" (n = 4,062)

Student-level variables (n = 6,099 students): Dependent variable: Reading achievement

Demographic variables:
Academic background
Percentage of minorities
Percentage of language minorities

Hope supports for literacy:
Parents' education (years)
Mother's expectation for child's

education (years) Public library use (personlfamily) Literary resources in home Discusses school with family factorc

School-level variables (n = 377 schools):

School composition and structure:
Percentage of poor children
Percentage of minority children
Percentage of urban location
School sector:

Percentage in public school Percentage in Catholic school Percentage in other private schools

Percentage of stand-alone middle school Eighth-grade enrollment Average achievement

School conditions: Home support for learning factorc Parents' satisfaction with schools factorc Average teacher absenteeism (days/

semester) Average student absenteeism (last month) Orderly environment factorc Positive teacher-student relations factorc Social capital factorc School safety factorc

School policies and practices:
Teacher cooperation factorc
Books assigned in English class

(nontextbooks) Authentic instruction in English class factorc Nongrouped class structure factorc

7.92 11.04***b

-.29 .14*** 49.57*** 22.92 23.00*** 8.54

12.17 13.66***

14.77 15.52***

1.19 1.48*** 4.66 5.82*** -.23 .12***

40.79*** 23.34
38.38*** 2 1.23

25.12 22.99

94.82*** 89.94

4.49 8.18*** .69 1.88**

67.67 66.42

212.64 208.56

9.27 10.58***

-.I4 .07*** .02 -.01

2.43 2.50 1.62*** 1.48

-.02 .01
.08*** -.03

1.92 2.08*** -.11 .05***

.O 1 -.O1

2.71 2.85**

.O 1 -.01 .06*** -.29

a Sample sizes are unweighted. Group mean differences were computed with NELS

design weights. This convention applies to all statistics reported in this article. Group mean differences were tested with t-tests. Factors were standardized on the student sample with mean = 0, SD = 1.

** P s .01.
*** P s,001.

membership), the two groups differ considerably on other measures of demographic conditions. Poor children have weaker academic back- grounds (.4 SD lower) and are more likely to be raciallethnic minority group members (almost 50 percent are black or Hispanic, as compared to 23 percent of the middle-class group). While close to a fourth of poor eighth graders are classified as language minorities, less than 10 percent of middle-class children are so classified (certainly because there are more Hispanics in the poor group).

On every home support measure considered, middle-class children are favored. Compared to poor children, those from middle-class fami- lies have more educated parents, have mothers who expect them to go farther in school, make more use of the public library, have more resources in the home that are related to literacy, and spend more time discussing matters related to school with their families.

Variables describing schools. -Middle-grade schools are highly strat- ified in economic, raciallethnic, and academic terms, as indicated by the higher proportion of poor and minority students in the schools attended by the poor children in our sample and by the lower average achievement in those schools. On several structural features, however, the schools attended by the two groups are similar: urban location (about 25 percent in both cases), grade structure (two-thirds of chil- dren in both groups attend stand-alone middle schools), and grade size (most attend schools with about 210 eighth graders). While the vast majority of children in the sample attend public school (90 percent or more), middle-class children are more likely to attend either Catholic or other private schools.

In general, school conditions are better for middle-class than for poor students. Middle-class children attend schools where principals describe more support from the home for learning, with more social capital, and are described by children and parents as having safer environments. Schools attended by poor students have higher levels of student absenteeism but have somewhat more positive relationships between teachers and students. The schools for these two groups do not differ on parents' satisfaction with the school, in average teacher absenteeism, or in their orderly environments. Mean differences in the schools' policies and practices are fewer for the two groups of students. Middle-class students' schools use more outside books in their English classes, whereas schools attended by poor students are more likely to have ungrouped classes. There are no differences between groups in teacher cooperation or in the prevalence of what we call authentic instruction in the English classes.

The substantial differences between poor and middle-class chil- dren's family and school demographic conditions indicate the strong

302 American Journal of Education need to control for these variables in multivariate analyses. The effects of differences in home and school supports for literacy-also substantial-constitute the focus of the remainder of our analyses. We first revisit the question, now within a multilevel context, of how home supports for literacy differentially affect the reading comprehension of young adolescents living in poverty or in middle-class circumstances.

Correlations between Model Variables

Within-school variables. -Bivariate relationships between all variables in our final HLM models, presented as zero-order correlations, are displayed in tables 2 and 3. Table 2 shows correlations among variables measured on individuals and their families. These correlations were computed with the use of the NELS student-level design weights. It is not surprising that the correlation between reading achievement and academic background is quite strong (r > .4),and academic back- ground is also strongly related to the mother's educational expecta- tions. Other strong correlations (all with r > .3)are between minority status and language minority status, between home literary resources and poverty status, and between literary resources and parents' educa- tion. Most other interrelations between these within-school variables are moderate (correlations between .2 and .3).There are several vari- ables with which both minority and language minority status are almost unrelated: the mother's expectation, academic background, and the frequency with which students discuss school matters with their fami- lies. These relationships are similar to what the extensive literature on family background and educational development has found.

Between-school variables.-In general, the strongest correlations are between school variables and average reading achievement (table 3). Notable among these are correlations between average achievement and minority concentration (r < -.5), school social capital, average student absences, and average number of books assigned (r > .4). Other moderately strong relationships exist between minority concen- tration and school social capital (r < -.4), average achievement and home support for learning, social capital and average student absen- teeism, and social capital and average number of books assigned (r > .3, r > -.3).Two variables describing instruction are only very modestly correlated with other variables in our model: authentic instruction and heterogeneous grouping. Variables describing teachers are also only very modestly related to other variables: teacher-student relations, teacher cooperation, and teacher absenteeism. On the basis of these bivariate correlations, it is clear that multivariate analyses are required

May 1994 303

Zero-Order Correlationsfor Within-School Model Variables"
Frequency of
Language        Mother's        Home    Discussions
Minority    Minority    Parents'    Educational    Academic    Literary    between Parents
Poverty    Status    Status    Education    Expectations    Background    Resources    and Students
Minority status        
Home literary        
" Within-school correlations were computed with the NELS student weights; N = 6,099.        

Zero-Order Correlations for Between-School Model Variables"
--    --    ~
Minority Authentic Concentration Instruction    Teacher Absenteeism    Teacher-Student Relations    Heterogeneous Teacher Grouping Cooperation    Support for Learning    School Social Capital    Student Absences    Number of Books Assigned
Average achievementb Minority concentration Authentic                
instruction Teacher absenteeism Teacher-student                
relations Heterogeneous grouping Teacher cooperation Home support for learning School social capital Student absenteeism                
"Between-school correlations used the NELS school weights; N = 377.                
To examine bivariate relationships between school variables and reading achievement, reading achievement was aggregated to the school level.

to sort out the unique relationships among the variables measuring both home and school supports and students' reading comprehension.

Multilevel Analyses for Home Support on Reading Achievement

Unconditional HLM model. -The analyses for this study that investi- gate both home and school supports for reading achievement employ the HLM methodology. The first step in an HLM model, before statisti- cal controls are introduced, is the partitioning of the total variability in each outcome into its within-school and between-school compo- nents. The effects of school supports may be evaluated only on the proportion of variability in the outcome that occurs between schools (the intraclass correlation). In the case of of reading achievement, this figure is 19.0 percent." Reliability of this outcome, estimated in HLM, is quite high (.715).

Within-school HLM model. -A two-step within-school HLM analysis investigating the effects of home supports on reading achievement is presented in table 4. All parameter estimates are presented as gamma coefficients in the metric of score points on the reading comprehension test. We are examining two 6 parameters as random effects (i.e., they are allowed to vary across schools): adjusted mean reading achieve- ment (Po) and the poverty gap on reading achievement (6,). Other within-school parameters are treated as fixed effects; thus, they are not examined as functions of school-level variables. Model 1 (see table 4) investigates the effect of poverty status on achievement. Mirroring results from table 1, we see that poor students' reading achievement is significantly below their middle-class counterparts' (an effect of

-2.753 points, or .5 SD). In this simple HLM model, poverty status alone explains over 13 percent of the between-school variance.

Model 2 (see table 4), which takes demographic and home support factors into account, reduces the poverty effect by over half.'' Adjusted school-mean reading achievement is 10.950, which is the score for white middle-class students in each school, pooled across the 377 mid- dle-grade schools in this sample. The poverty gap (-1.135) indicates the average number of points below white middle-class children that poor children scored in this multivariate model (about .2 SD lower). Academic background has an especially large effect (coefficient of 2.241), which supports our use of this measure as a proxy for ability. RaciaVethnic minorities ( -.817) and language minority students also score significantly lower (-.948), even after poverty status and aca- demic background are taken into account. Home support effects are also substantial. These include the effects of parent resources (their

306 American Journal of Education


Within-School Hierarchical Linear Model Estimating Home Supportsfor Reading Achievement



Random effects: Mean reading achievement (Po) Poverty gap (P,)

Fixed effects:

Demographic controls: Minority status Language minority status Academic background

Home supports for literacy: Parents' education Mother's educational expectation Literacy resources in the home Discussion of school matters with

familv Proportion bf variance explainedb


Model 1 Model 2'

Average achievement .567 .460 Poverty gap .lo0 .084


Estimated        Chi-
Parameter Variance    Degrees of Freedom    Square Statistic
Average achievement Poverty gap            
"All estimates for two-level models reported in this article were computed using the HLM program (Bryk et al. 1989).

This is computed as the decrease in estimated between-school parameter variance from the fully unconditional model (4.892, in the case of reading achievement) which is attained by this model.

'The chi-square statistics provide only approximate probability values. See Bryk and Raudenbush (1992) for more detail. *** P s .001.

education and expectations), as well as those for home literacy re- sources and student-family discussions about school. Together with poverty status, demographic and home support factors account for a substantial 55 percent of the between-school variance in reading achievement.

Hierarchical linear modeling computes reliabilities for the two ran- dom effects. Reliability of the average achievement measure is accept- able (around .5 for both models), although the reliability of the poverty gap is very modest (around .09).13 However, the chi-square table for the HLM model 2 in table 4 suggests that, while a substantial proportion of the between-school variability in the random effects is explained by this within-school model, there is adequate residual parameter variance in both random parameters to proceed (i.e., the chi-squares are statisti- cally significant).

Final HLM model. -The fully multivariate model displayed in table 5 investigates school effects on the two random /.3 parameters estimated in the within-school model from table 4: mean reading achievement and the poverty gap. These analyses include adjustment for the full set of demographic and home support variables analyzed in model 2 of table 4. We arrived at the final model by using the variable selection- retention strategy described previously.

For mean reading achievement, only one variable describing school composition and structure was retained in the "intercept" model (i.e., the HLM analysis of mean reading achievement). Somewhat surpris- ing to us is the lack of association with several school variables com- monly found to be important in school effects research. An urban location, poverty concentration, school sector, grade grouping, and grade size are not related to mean reading achievement.I4 However, schools with more minority students have significantly lower achieve- ment. In terms of school conditions, we found that schools receiving more support for learning from the students' homes have higher achievement, while schools where English teachers are absent more often have lower achievement. Two variables describing policies and practices are also positively related to average reading achievement: schools where teachers cooperate and coordinate more and those where more books (other than the text) are used in English classes. On the other hand, schools that group fewer of their classes by ability are those with lower achievement. The full HLM model explains 68 percent of the between-school variability in reading achievement, which represents an improvement of 13 percent and a proportional gain in explanatory power of 24 percent.

Because the poverty gap has a negative slope (adjusted mean gap of -1.029), variables that are positively related to this outcome induce

308 American Journal of Education

Full Between-School Hierarchical Linear Model Estimating School Supports fm Reading Achievement"

Random Effect Gamma Coefficienta

Mean reading achievement: Mean School composition:

Percentage of minority enrollment

School conditions: Home support for learning Average teacher absenteeism

School policies: Teacher cooperation Books assigned in English class Nongrouped class structure

Poverty gap: Mean poverty gap School conditions:

Home support for learning Positive teacher-student relations Social capital Average student absenteeism

School policies: Authentic instruction in English class Nongrouped class structure

Proportion of variance explained:b Mean achievement Poverty gap


Estimated Parameter Variance    Degrees of Freedom    Chi-Square Statistic
Average achievement Poverty gap            
a This model includes all within-school controls shown in table 4.

This is computed as the decrease in estimated between-school parameter variance from the unconditional model (4.892 in the case of mean achievement; 1.00712 for the poverty gap) attained by this model.

* P 6.05.

** P 6.01.

*** P 6 .001.

equity between poor and middle-class children. We described earlier the characteristics of effective schools as those that were simultaneously and positively related to both the intercept and the slope. No single school characteristic satisfies that criterion. However, two variables-home support for learning and nongrouped class struc- ture-are related to both the intercept and the slope, but in opposite directions. In the case of home supports, higher levels of parental involvement in learning tasks foster higher mean reading achieve- ments, but they also induce greater differentiation in this outcome between poor and middle-class students. Because this type of support comes more often from middle-class than poor parents, this finding is disturbing but not surprising. In the case of heterogeneous grouping, differentiating classes by ability is associated with higher mean achieve- ment, as well as greater stratification in achievement between students in the same schools. Given what we know about tracking, the effects on mean achievement are troubling, but the effects on achievement stratification are quite consistent with the literature on this topic (see, e.g., Knapp and Turnbull 1990; Lee et al. 1993; Oakes 1985). Other school conditions that are positively related to the slope (i.e., those that induce equity) include positive teacher-student relations and higher social capital. High levels of student absenteeism induce social inequality in achievement.

Among the most important findings in this study, in our opinion, are those concerning instructional policies and practices. Certain of these policies-in particular, authentic instruction in English classes and the prevalence of nongrouped classes (in all subject areas, not just English)-lead to social equity in achievement. These findings seem reasonable and logical. While the effects of ability grouping on equity have been explored (Gamoran 1991; Lee and Smith 1993), we are aware of no studies that link authentic instruction to educational eq- uity. As the poverty gap is estimated with low reliability (typical of slope estimates), the proportion of explained variance in this outcome is low- 10 percent.


Reducing Inequities

Inequity in the literacy development of young adolescents that results from their growing up in families defined by different economic condi- tions constitutes a major conceptual focus of this study. Our purpose

3 10 American Journal of Education has been to identify conditions in the home and school environments of these two groups of children-particularly conditions that are sub- ject to alteration-that are associated with a reduced learning disad- vantage for poor eighth graders (what we've called the poverty gap). At least in part, the study has accomplished that objective. This process is shown graphically in figure 1. The numbers used to construct these graphs are taken directly from HLM results, mostly those from tables 4 and 5.15

The graph displays the poverty gap in reading achievement under three analysis conditions: observed (i.e., unadjusted; the solid line), the taking into account of differences in demographic background and home supports for literacy between the two groups (the dashed line), and demographic, home, and school supports for literacy accounted for (the dotted line). The "story" conveyed by this graph is as follows. If the home and school supports identified here were somehow made equivalent for young adolescents living in poverty and their middle- class counterparts, the disparity in reading achievement levels between groups would diminish substantially. This is demonstrated by the sys- tematic reduction in each line's slope as conditions of home and school are taken into account. However, even if these conditions were com- pletely equalized, the achievement disparity would not disappear com- pletely (i.e., the dotted line still has a negative slope).

Although our findings suggest that changing the home and school

conditions for poor children would substantially improve their

achievement in reading, which factors are most important? It is our

desire to help the reader derive substantive meaning from the statistical

results of a set of somewhat complex analyses. While we believe that

the results from this study have implications for educational policy,

we also exercise some caution in approaching their interpretation.

The study employs cross-sectional data and correlation-based meth-

ods. With these cautions in mind, we turn to a general discussion of

home and school conditions and how they might be changed.

Home, Families, and Reading

Conventional wisdom holds that, while school policies and practices may be changed, home conditions are immutable. On the basis, at least in part, of our findings here, we cannot support this argument. Many of the home supports we have considered in this study are subject to change, and much of the impetus for change can (and should) come from the school. While it is difficult to advocate a policy of raising parents' educational levels (although training in adult liter-

acy as a social intervention is not only possible but happening in some settings), urging parents to hold high expectations for their children is something schools can certainly do. As Lee et al. (1993) state in a sum- mary of the literature on this issue, "Parental expectations for their children's achievement and the importance that parents place on educa- tion are positively and strongly related to academic performance" (p. 20).

Our findings about home support for literacy are consistent. The presence of reading material in the home-whether borrowed from the library or owned (the families' own books, newspapers, or maga- zines)-is directly associated with children's literacy development in terms of their reading comprehension. This finding offers more gen- eralizable empirical support for the ethnographic analysis by Snow et al. (1991) of a small sample of poor families. Schools are certainly able to assist parents of all social backgrounds in obtaining reading materials. Urging students to make use of the public library for assign- ments, encouraging parents to provide reading materials for their children, and emphasizing the importance of such things to parents is within the capability (perhaps the responsibility) of all schools.

Our findings also suggest that family discussions about school expe- riences and educational plans contribute to reading comprehension. Schools can facilitate these discussions by providing parents with infor- mation and activities that require parent-student interactions. Poor families might especially benefit from support and encouragement of this type. Moles (1987) found that poor parents would like more input from their children's schools, and they were unhappy that the major school contact was to report "bad news." Studies have found that schools can successfully promote parent-student interactions about school experiences and plans, and that these efforts contribute to gains in students' achievement (Ascher 1988; Epstein 199 1).

Social Composition of Schools

Findings here support a large body of research that documents the effects of increasing social stratification in educational outcomes by concentrating poor and minority children in some schools. There is a body of literature, reviewed by Mahard and Crain (1983), that docu- ments the positive effects of racial integration of schools on student achievement. Our results offer indirect support for these conclusions. Given the history of desegregation in this country and the disappoint- ments that surround it, we are not sanguine about the possibility of dramatically changing the racial composition of schools (Bates 1990; Vergon 1990).

School Social Conditions

Two findings offer some support for the notion that schools where positive social and professional relations are the norm are environ- ments where more learning takes place. The first finding focuses on policy, where higher levels of cooperation and coordination among teachers are associated with higher average reading achievement. Such schools are more likely to use teacher teaming (within the same subject or with interdisciplinary subjects), as well as to provide for common planning time and/or flexible scheduling. The second finding relates to school social interactions, whereby schools with positive social relations between students and teachers are those in which achievement is more likely to be equitably distributed. Although a measure of schools, re- ports of these relationships come from students, which is an important point in supporting the validity of the measure. These findings, when taken together, suggest that social relations among and between stu- dents and staff are important factors in learning-beyond home in- fluences and the composition of students.


The topic of school absenteeism has been subjected to little empirical scrutiny. What research exists focuses on students, especially on their behaviors as individuals, and concentrates on absenteeism as a precur- sor of dropping out (Bryk and Thum 1989). In this study, we focus on absenteeism as a school phenomenon rather than as an individual phenomenon. Moreover, we have considered teachers' (particularly En- glish teachers') absences as well as those of students. The coefficient for teacher absenteeism suggests that schools in which English teachers are absent more often are those where students learn less. An alterna- tive interpretation might be that, in schools where achievement is low, teachers are less committed. It is interesting that, while teacher absenteeism is associated with average achievement, student absenteeism is related to equity. That is, more student absenteeism widens the poverty gap in achievement. Such a result is probably explained by the fact that it is poor children who are absent more often.'Wur interpretation of the results on absenteeism is as follows: when teach- ers are absent, all students are affected, but, when students are absent, it is mainly the absentees themselves who suffer. Our findings about absenteeism warrant closer scrutiny, particularly in pinning down the causal direction of achievement and absenteeism. We are pursuing this topic in another study.

314 American Journal of Education

The Home and the School

Home support for learning.-Besides the home supports already dis- cussed (library use, books, magazines, and the like), our results provide more empirical evidence to the vast body of research in the area of parental support for learning. The Carnegie Council on Adolescent Development (1989) urges schools to "reengage families in the education of young adolescents . . . [by] offering families opportunities to support the learning process at home and at the school" (p. 9). The sort of home support described by the variables in our study also includes the notion of strong school-family ties, which surely result in a closer social align- ment of values between home and school (Comer 1980, 1988).

While our findings underline the positive association of home sup- port for learning with mean levels of reading comprehension, high levels of such support in schools also appear to have a cost in that they appear to promote inequity in the social distribution of reading achievement. We know that higher levels of home support typify schools attended by middle-class children and that middle-class parents have more conversations about school experiences and plans with their children. This suggests that the support is actually coming from mid- dle-class families rather than from those living in poverty-a finding well documented in the literature (Epstein 1985; Grant and Sleeter 1988).17 While schools need to encourage parents to involve them- selves in the school and in their children's learning, our findings sug- gest that more effort should be directed toward involving the children of poor parents, who typically wait for an approach from the school to become involved. As children advance in age and grade, moreover, such parents may have increasing difficulty actually helping their chil- dren with their school assignments. This suggests possible school inter- vention at a very concrete level-to help parents of limited education learn along with their children.

Social capital. -We believe that our measure of school-level inter- family contact (i.e., parents knowing their children's school friends and the parents of those friends) taps the construct of social capital on which schools may draw, which is an idea that is convincingly discussed by Coleman (1987, 1988). Aggregate results here suggest that social capital induces social equity, in that schools with higher social capital are those with a lower poverty gap in achievement. We also note that the relationship between social capital and average read- ing comprehension was positive and significant in a full HLM analytic model of school conditions, before variables measuring instructional policies and practices were introduced. The strength of bivariate corre-

May 1994 315 lations between social capital and other variables in our model (see tables 2 and 3) suggests the importance of this construct in students' academic lives.

Classroom Conditions

More books.-We suggest that our most interesting findings are re- lated to variables tapping the policies and practices that affect eighth- grade English classrooms. Most straightforward is the finding that if English teachers assign more books to read in their classes (in addition to the text), their students are likely to better comprehend what they have read. This finding is consistent with those discussed above about books and reading material in the home. Quite reasonably, when stu- dents are exposed to more books, whether in their English classes or in their homes, whether obligatory (in school) or optional (at home), they seem to read better.

Authentic instruction.-We suggest that our measure, labeled "au- thentic instruction in English classes," actually taps English teachers' efforts to encourage students to construct (rather than reproduce) knowledge. The centrality of student writing, including editing that writing, forms an important component of the composite measure, which also includes items querying the use of literature and oral pre- sentations. Our evidence suggests that if poor children were exposed to more "authentic" instruction in English, their reading comprehension would increase.18 Although there is evidence in NELS and elsewhere (Newmann 1992) that instruction of this sort is rare in American schools, it is also clear that low-ability children (who are all too often poor children) are least frequently taught in this way (see, e.g., Knapp and Needels 1990; Knapp and Turnbull 1990; Oakes 1985). More often, classes containing high proportions of such children involve large amounts of drill and practice on basic skills.

Although the finding that this variable has no independent influence on mean achievement suggests no "cost" in implementing such a policy (i.e., overall achievement would be unaffected), we find the lack of an independent effect here somewhat disturbing. Although the relation- ship of authentic instruction with average achievement was positive, its magnitude did not achieve statistical significance. Advocates of au- thentic instruction assume that all children will learn in classrooms that offer this form of teaching, not just children from poor families. A reasonable explanation for the "no difference" finding here is that our measure of this construct is somewhat "blunt," in that the effects of authentic instruction should be evaluated in classrooms. rather than

3 16 American Journal of Education in schools (unfortunately, classroom-level analyses are not possible with NELS data). We encourage other researchers to pursue this issue within and across classrooms. Given the character of this variable, our findings probably represent a lower bound for the true relationship between authentic instruction and student learning.Ig

Decreased ability grouping. -An issue of "cost-benefit" surrounds the findings about another school policy-heterogeneously grouped in- struction across subject areas (synonymous with what some call "de- tracking"). In this instance, while nongrouping is positively associated with the equitable distribution of reading achievement (i.e., it decreases the poverty gap), nongrouped class structure is simultaneously nega- tively associated with school mean achievement. Some might conclude that, as implemented in America's middle-grade schools in 1988, het- erogeneous grouping may be seen as narrowing the poverty gap in a way that suggests that everyone does worse in such schools-poor and middle-class children. We disagree with this interpretationz0 and strongly urge that these findings not be interpreted as encouraging more tracking in middle-grade schools. In fact, we agree with the Carnegie Council on Adolescent Development's (1989) conclusion that "tracking has proven to be one of the most divisive and damaging school practices in existence" (p. 49), especially for young adolescents.

Rather, we suggest that a policy where schools blindly implement detracking without some accommodation for differences in students' skill levels could be accounting for these findings-everyone would lose under those circumstances. Instead, we encourage a remediation policy in which a school's resources might be redistributed rather differently from the current practice. Rather than grouping together in one class students whose skills are weak (here, we mean skill in reading) in special classes labeled "remedial," "basic," "low ability," or "developmental," we encourage a policy where students are placed in regular classes without regard to ability level, but where those students whose skills are weak receive extra instruction in addition to and coor- dinated with their regular classroom instruction." This policy would redirect school resources toward those students most in need of them. This approach to remediation goes far beyond federally funded com- pensatory programs like Chapter 1. These results hint that simple detracking as presently practiced in America's middle-grade schools, without a reconstructed approach to remediation, may be misguided.

What Doesn't Count

In building our final HLM models, we considered a number of vari- ables that have been the focus of many school effects studies (including our own). Many of the variables we considered that define schools in terms of their structure had little influence here. We are loath to draw substantive conclusions from nonsignificant findings because of the potential of large Type I1 error rates. However, given our focus on family poverty, we were somewhat surprised that such factors as urban location or governance structure (i.e., school sector) were unrelated to either average reading comprehension or its social distributi~n.'~ Compositional factors also did not show a strong residual influence (except minority concentration), after school conditions, policies, and practices were taken into account.

Two explanations for the lack of importance of structural features seem reasonable. One is related to the sample considered here. Since our sampling strategy led us to eliminate the top quartile of students, defined in terms of adjusted family income, it may be that much of the variability in structural effects on student performance comes with the inclusion of affluent students, who are more likely to attend private schools and unlikely to live in large cities. A second explanation is substantive. While school effects research has typically focused on school structure and ignored instruction, here we have attempted to coordinate our investigation of students' skills and behaviors in reading with several variables defining their teachers and classrooms in the same subject area. Although they were writing on a different topic, we agree with the conclusions of Bryk and Lee (1992) in this regard, who state: "Student learning is not like the Gross National Product. There is no theoretical justification for a concept of 'total student achievement.' . . . Students learn individual subjects especially when they are taught those subjects" (p. 441). Later, they conclude that "school effects re- search has increasingly defined subject matter as a factor that must be included in the study design" (p. 449).

Schools and Classrooms Do Count

We conclude by reiterating that America's middle-grade schools have major opportunities and responsibilities for equalizing the literacy development for all of their students, regardless of the economic condi- tions of the students' families. It is too easy for schools to ascribe the learning disadvantages of their less affluent students to deficient home environments. While this study has underscored that such homes are deficient in terms of reading materials and opportunities, our results indicate that schools should act to correct these deficits, both through encouraging parent involvement in the school and supporting parent engagement in their children's reading activities.

3 18 American Journal of Education

Even more important, and a somewhat less controversial activity for schools, is how they structure instruction in English and reading in the middle grades. While we support a move away from homogeneous instruction, we also encourage a simultaneous approach to remedia- tion that directs more school resources to children most in need. In- struction in English classes that involves writing, editing, oral presenta- tions, and a healthy dose of literature is shown here to "close the poverty gap." Moreover, requiring students to read more books in- duces more learning as well.

In addition to what is taught and how that instruction is delivered, our findings suggest that teachers are also important. Schools where teachers are given an opportunity to work together, where they coop- erate and coordinate their teaching with other teachers, and schools where teachers are absent less frequently seem to be places where students learn more. Schools where teachers have positive relations with their students also appear to be more equitable environments.

There is a consistent thread running through the results of this study. Middle-grade schools where learning is undifferentiated by abil- ity or social background, where high-level instruction is the norm, where students and teachers are socially engaged in cooperative en- deavors toward learning-these seem to be schools with high levels of literacy development and where learning is distributed equitably. These are schools that are typified by notions currently under active discussion in the reform literature as "communally organized" (Lee et al. 1993) or "restructured" (Lee and Smith 1993). In sum, our findings lend some empirical support for reform efforts in these directions.

Appendix Description of Variables

Means and SDs are unweighted. Weighted values are used for analyses. Con- tinuous variables are standardized with a mean of 0 and an SD of 1 in HLM analyses.


Dependent Variable

Reading achievement. -Data are the IRT-estimated formula scores on a test of reading comprehension (BYTXRIRS). Scores range from -0.63 to 20.86 with a mean of 10.03 and an SD of 5.84.

Demographic and Other Control Variables

Poverty status.-Data are the ratios of family income (BYFAMINC) and family size (BYFAMSIZ) to 1987 federal poverty standards. Data are dummy coded so that 1 is poor (the lowest quartile) and 0 is middle income (the second and third quartiles of the ratio). Thirty-two percent (32.3%) of the families in the study's sample are poor.

Minority status. -Data are dummy coded so that 1 signifies a black, Hispanic, or Native American student and 0 signifies a white or Asian student (BYS3 1A). Thirty-two percent (31.5%) of the students are minorities.

Language mznorzty status. -Data are dummy coded so that 1 signifies a lan- guage minority and 0 signifies a language majority (BYLM). Thirteen percent (13.2%) are language minority.

Academzc backgrrmnd. -Data are averages of composites of self-reported grades in science, math, English, and social science for grades 6-8 (BYGRADS) and parents' reports of whether students ever repeated a grade (BYP44). Values range from 0 to 2.50 with a mean of 1.3 and an SD of 0.39.

Home Supports for Literacy

Parents' education.-Data are the highest number of years of education obtained by the mother, father, or guardian (BYPlAl, BYPlA2, BYP30, and BYP31). Student reports were used for missing parental data (BYS34A and BYS34B). The number of years ranges from 6 to 20 with a mean of 13.20 and an SD of 2.54.

Mother5 expectations for child's education.-Data are the number of years of education that the mother or female guardian expects the child to obtain (BYPlAl, BYPlA2, and BYP76). Student reports were used for missing paren- tal data (BYS48B). Values range from 10 to 20 with a mean of 15.28 and an SD of 2.32.

Public libra~ use.-Data are the sum of two dummy-coded variables (1 = yes; 0 = no) for whether the parent borrows books from the public library (BYP61AA) and whether the student borrows books from the public library (BYP61AB). Values range from 0 to 2 with a mean of 1.39 and an SD of 0.76.

Literacy resources in the home.-Data are the sum of eight dummy-coded variables (1 = yes; 0 = no) for whether the family gets a daily newspaper (BYS35B), subscribes to or regularly purchases a magazine (BYS35C), and owns an encyclopedia (BYS35D), an atlas (BYS35E), a dictionary (BYS35F), a typewriter (BYS35G), more than 50 books (BYS35M), and a home computer that can be used for schoolwork (BYP70). Values range from 0 to 8 with a mean of 5.45 and an SD of 1.74.

Family discussions about school ,factor.-Data are standardized factor scores that have been computed by principal components factor analysis. Scores include how often the student reports discussing school programs with the parents (BYS36A), school activities with the parents (BYS36B), and classroom studies with the parents (BYS36C); how often the student reports talking about high school plans with the father (BYSSOA), the mother (BYSOB), and the relatives (BYS503); and how often the parents report talking to the student

320 American Journal of Education

about school experiences (BYP66), high school plans (BYP67), and plans after high school (BYP68). The eigenvalue is 2.85. The percentage of variance is 31.7%. Alpha is .72.


School Composition and Structure

Percentage of poor children.-Data represent poverty status aggregated to the school level with use of the full sample (POOR). Percentages range from 2 to 93 with a mean of 32.5 and an SD of 19.3.

Percentage ofminority children. -Data are the percentages of black, Hispanic, and Native American eighth graders enrolled at each school. Percentages are summed with the use of privileged data (BYSClSAP, BYSClSCP, and BYSC13DP). Percentages range from 0 to 100 with a mean of 25.37 and an SD of 31.55.

Urban environment. -Data are dummy coded (1 = yes; 0 = no) for whether the school is located in an urban center (G8URBAN). Twenty-one percent (20.7%) of the schools are urban.

School sector.-(A) Catholic, data are dummy coded (1 = yes; 0 = no) for whether the school is Catholic. Nine percent (8.5%) of the schools are Catholic (GSCTRL). (B) Other private, data are dummy coded (1 = yes; 0 = no) for whether the school is non-Catholic and private. Five percent (4.9%) of the schools are private non-Catholic.

Stand-alone middle school.-Data are dummy coded (1 = yes; 0 = no) for whether the school is limited to a combination of grades 6-9 (GRADSPAN). Fifty-seven percent (57.3%) of the schools are middle schools.

Eighth-grade enrollment.-Data are the natural log of the school's eighth- grade enrollment. Log values are calculated with the use of a continuous version of the variable on the privileged data file (G8ENROLP). Values range from 2.40 to 7.19 with a mean of 4.73 and an SD of 0.91.

School Conditions

Home support for learningfactor. -Data are standardized factor scores calcu- lated with principal components factor analysis. The learning factor includes the percentage of the parents administrators believe (a) monitor homework (HES26D), (6) attend parent-teacher conferences (HES26G), and (c) encour- age learning in the home (HES26H); the average number of times parents say they contacted the school about their child's academic performance (BYP57A); and the percentage of students who say their parents spoke to a teacher or counselor during the year (BYP37B). The eigenvalue is 2.93. The percentage of variance is 58.6%. Alpha is .82.

Parents' satisfaction factor. -Data are standardized factor scores calculated with principal components factor analysis. The parents' satisfaction factor includes the degree to which parents believe learning is a high priority at school (BYP74A), homework is worthwhile (BYP74B), their child is challenged

May 1994 321

(BYP74C), their child works hard at school (BYP74D), their child enjoys school (BYP74E), school standards are realistic (BYP74F), school prepares students for high school (BYP74G), school prepares students for college (BYP74H), and their satisfaction with the school (BYS75). The eigenvalue is 5.95. The percentage of variance is 66.1%. Alpha is .93.

Average English teacher absenteeism. -Data are the average number of days English teachers reported being absent during the first semester of the year. Values aggregated from the full sample to the school level (BYT3-28). Aver- ages range from 0 to 12 days with a mean of 1.97 and an SD of 1.67.

Average student absenteeism.-Data are the average number of days eighth graders were absent in four weeks. Values are aggregated from the full sample to the school level (BYS75). Averages range from 0 to 4.83 days with a mean of 1.47 and an SD of 0.64.

Orderly en7nronmentfactor. -Data are standardized factor scores calculated with principal components factor analysis. The orderly environment factor includes the degree to which administrators think that discipline is emphasized at the school (BYSC47B), that classroom environment is structured (BYSC47D), that teachers encourage students to do their best (BYSC473), that students are ex- pected to do homework (BYSC47F), that the school day is structured (BYSC47J), and that deviation from school rules is not tolerated (BYSC47K). The eigenvalue is 3.67. The percentage of variance is 61.0%. Alpha is 37.

Positive teacher-studat relativnsfactm. -Data are standardized factor scores calcu- lated with principal components factor analysis. The teacher-student relations factor includes the degree to which students say that they get along with teachers (BYS59A), that teaching is good (BYS59F), that teachers are interested in them (BYS59G), that teachers praise their work (BYS59H), that teachers put them down (BYS59I), and that teachers listen to what they say (BYS59J). The eigen- value is 4.02. The percentage of variance is 66.9%. Alpha is .90.

Average social capital. -Data are the average number of parents an eighth grader's school friends' parents say they know (BYP62A1, BYP62B1, BYP62A2, BYP62B2, BYP62A3, BYP62B3, BYP62A4, BYP62B4, BYP62A5, and BYP62B5). Values are aggregated from full sample to school level. Aver- ages range from 0.25 to 4.40 with a mean of 2.20 and an SD of 0.73.

Safe schoolfactor. -Data are standardized factor scores calculated with princi- pal components factors analysis. The safe school factor includes the average number of times students report being offered drugs at school (BYS57B), the degree to which students believe that physical conflict is a problem (BYS58D), that vandalism is a problem (BYS58F), that alcohol use is a problem (BYS58C), that drug use is a problem (BYS58H), and that weapons are a problem (BYS58H), the extent to which students do not feel safe at school (BYS59K), the extent to which parents think the school is a safe place (BYP74I), and the degree to which English teachers believe that physical conflict is a problem at school (BYT3-26D), that robbery is a problem (BYT3_26E), that vandalism is a problem (BYT3_26F), and that weapons are a problem (BYT3-261). Variable direction is reversed. The eigenvalue is 5.99. The percentage of variance is 49.9%. Alpha is .90.

School Policies and Practices

Average number of books assigned in English classes.-Data are the average number of books English teachers assign during the year in addition to text-

322 American Journal of Education book or workbook selections. Values are aggregated from full sample to school level (BYT2-19). Averages range from 0 to 5 or more books with a mean of

2.79 and an SD of 1.79.

Authentic instruction in English classes factor-(encouragement of students to construct knowledge). Data are standardized factor scores calculated with principal components factor analysis. The authentic instruction factor includes administrator reports of how frequently the typical English class (a) requires students to write reports of at least one page (HES198AA), (b) teaches content and ideas in works of literature (HESlgAC), (c) requires students to rewrite and resubmit compositions after peer or teacher review (HESlgAD), and (d) has students make oral presentations (HES19AE). The eigenvalue is 2.09. The percentage of variance is 52.3%. Alpha is 68.

Heterogeneously grouped classfactor. -Data are standardized scores calculated with principal components factor analysis. The grouped class factor includes whether the school has grouped academic classes (HES7Cl); a composite indicating that 40 percent or less of sampled students report being in non- grouped math classes (BYSGOA), science classes (BYSGOB), English classes (BYSGOC), and social studies classes (BYSGOD); and whether eighth graders change classmates for different classes (HESGC). The eigenvalue is 1.62. The percentage of variance is 54.2%. Alpha is .49.

Teacher coordination and cooperation factor.-Data are standardized factor scores calculated with principal components factor analysis. The teacher coor- dination and cooperation factor includes whether interdisciplinary teams of teachers share the same students (HES23G2), the class periods are flexible in length (HES23H2), the eighth grade uses departmental team teaching (HES27C) or interdisciplinary team teaching (HES28C), and there is sched- uled common planning time for departmental members (HES2312). The ei- genvalue is 1.65. The percentage of variance is 55.0%. Alpha is 55.


Resnick (1991) refers to two other forms of literacy: literacy for pleasure (the use of print for personal enjoyment or leisure) and useful literacy (the use of print to mediate action). This latter form includes what has been de- scribed as functional literacy as well as more advanced skills associated with effective citizenship and employment. Each form has also been the focus of public campaigns and school-based interventions (see, e.g., Committee for Economic Development 1987).
Our original analyses included a full investigation of a second outcome variable-students' self-reported time per week spent on reading for pleasure. Because of the limited reliability of this single-item measure, and because our results were rather similar to those reported for reading comprehension, we have not reported these results in this article. Full details of the analyses on time spent reading for pleasure are available from the authors.
Our discussion of the literature on school effects is selective. A more complete treatment of this topic can be found in Lee et al. (1993).
The source of many of our school measures is a separate data file col- lected on NELS:88 schools by researchers at the Center for Research on Effective Schooling for Disadvantaged Students at Johns Hopkins University
May 1994 323

(Epstein et al. 1991). The data are in a public-use data file available from the National Center for Education Statistics.

For a family of four, this is equivalent to an annual income of less than $16,000. Thus, all students eligible for free lunches and approximately 70 percent of those eligible for reduced-cost lunches are included in our target group. See Department of Health and Human Services (1987) for a description of the poverty thresholds we used.
We use the academic background measure as a proxy for ability, since the base year data of NELS.88 are cross-sectional. Although it could be argued that this taps a concomitant measure of achievement (i.e., self-reported grades), the fact that students were asked to summarize their grades since the sixth grade gives the measure a retrospective tone. We recognize that we are making a compromise between potentially overcontrolling for academic status and overesti- mating home and school effects, because our dependent variable is measuring cumulative achievement without a pretest measure. We prefer to err on the side of overcontrol. In this decision, we are following the lead of several other school- effects studies in employing the academic background measure in this way (see, e.g., Lee and Bryk [1988, 19891; Lee and Smith [1993], who employ the same measure in a study that also employs base-year NELS data).
Independent variables (X's) may be specified as being of two types in HLM-random or fixed. It is typical that random variables are those whose coefficients we wish to investigate as "slope" outcomes in between-unit models (e.g., our poverty gap). On the other hand, fixed effects are relationships with variables that are important statistical controls but are those that we are not interested in modeling. The technical meaning of fixed effects in HLM is that their variance is constrained to be entirely within schools and is set to zero between schools.
While we might also wish for weak differentiating effects with respect to minority status, language minority status, and academic background (i.e., small values for and P4),these relationships are not at issue in this study and will not be discussed.
The statistical tests in table 1 do not reflect the hierarchical nature of our data, so some caution should be exercised when interpreting the P values. We use them only to describe our study sample and not to draw substantive conclusions about population differences.
We use the standards described by Rosenthal and Rosnow (1984) for effect sizes: effects below .2 SD are small, those below .5 are medium, and those above .5 are large.
The variability between schools (7)for reading achievement is 4.892. The variabilitv within schools, pooled across schools (a2),is 29.1967. When the latter figure is adjusted by the variable's HLM-estimated reliability (.715), the intraclass correlation is computed as (4.892)1[4.892 + (29.1967)(.715)].
We decided not to estimate the effects of demographic controls and home support for literacy in two separate models. While in this instance we have characterized parents' education as a home support (for reasons spelled out above), this measure is traditionally a major component of socioeconomic status and is, thus, a demographic variable. Although we have described these variables in separate sets, in fact we recognize the lack of a clear distinction between the two sets.
Low reliability is typical for slope outcomes (Lee and Brvk 1989), which suggests that most of'the observed variability in this slope is sampling variance and is, thus, unexplai~lable by school factors.
It should be noted that poverty concentration was significantly related to mean achievement in HLM models that included only composition variables. The addition of variables measuring school conditions, however, meant that this variable's effect fell to nonsignificance.
The figures for the unadjusted slope (solid line) were computed with HLM, where the only within-school variable was poverty status (model 1 in table 4). This results in a middle-class mean of 11.214 and a poverty effect of -2.753. Thus, the mean achievement of poor children is 8.461 (1 1.214 2.753). Figures for the means adjusted for demographics and home supports (dashed line) come from model 2 in table 3-a middle-class mean of 10.950, and a poverty mean of 9.815 (10.950 -1.135). The means for the line adjusted for demographic, home, and school supports (dotted line) use figures from table 4, which reports an adjusted middle-class mean of 11.081 (mean reading achievement plus the sum of related school effects, 10.815 + .266) and an adjusted poverty effect of -0.643 (the poverty effect as the sum of the school effects plus the mean poverty gap of -1.029). Thus, the mean achievement for children in Dovertv is 10.438.
324 American Journal of Education


In fact, poor children are absent significantly more often than middle- class children. They missed 1.9 days of school in the last four weeks, compared to 1.4 days for middle-class children (t = 7.2, P < ,001).
There are several indications of statistically significant differences among individual families in the NELS data on parental support for schooling. For example, middle-class parents report contacting schools more often to check on their children's academic performance, helping their children with their homework, and visiting their child's class more often than poor parents. Table 1 indicates a significant difference in how much poor and middle-class families talk about school experiences and plans.
An alternative interpretation of the influence of this measure, which we have labeled "authentic instruction," is that it may in fact be measuring a group of "compliance factors" of principals reporting about the frequency of certain desirable practices by English teachers in their schools. Principals who are aware of the currency of particular reform efforts may overreport these practices, and these may also be schools that enroll higher-achieving poor students. Of course, we favor the interpretation we have provided in the text.
As tables 2 and 3 show, our measure of heterogeneous grouping is uncorre- lated with minority status (r = .03) or average number of books read (r = -.04), two other school measures associated with average achievement in our model. It is significantly correlated with social capital (r = .27) and home support for learning (r = .22). In fact, most studies find grouping effects on achievement to be neutral on average, with significant negative effects on within-school distribu- tions of learning (Lee et al. [I9931 review these studies).
Investigating the prevalence of authentic instruction in classrooms, as well as its effects on students' engagement with learning, is a major undertak- ing of the National Center on School Organization and Restructuring at the University of Wisconsin-Madison and is sponsored by the U.S. Department of Education. Although one of us is a principal investigator with the center, the study described here was not conducted under the center's auspices.
2 1. Our views on an appropriate approach to remediation are very similar to those of Robert Slavin and his associates in the "Success for All" program in the lower grades (Slavin et al. 1990).

22. In our model-building, we considered such structural features as urban location and school sector before introducing school conditions and policies.

May 1994 325

Since these conditions and policies differ by urban location and sector, if the structural features were not investigated independently of these variable sets, one set might very well "explain away" the other. In fact, this did not occur.


Allington, R. L. "Effective Literacy Instruction for At-Risk Children." In Better Schooling for the Children of Poverty: Alternatives to Conventional Wisdom. Vol. 2, Commissioned Papers and Literature Review, edited by M. S. Knapp and

P. M. Shields. Washington, D.C.: U.S. Department of Education, Office of Planning, Budget, and Evaluation, 1990.

Applebee, A. N., J. A. Langer, and I. V. S. Mullis. Factors Related to Reading Achievement in Grades 3, 7, and 11. Princeton, N.J.: National Assessment for Education Progress, Educational Testing Center, 1988.

Ascher, C. "Improving the School-Home Connection for Poor and Minority Urban Students." Urban Review 20 (1988): 109-21.

Baker, D. P., and D. L. Stevenson. "Mothers' Strategies for Children's School Achievement: Managing the Transition to High School. Sociology ofEduca- tion 59 (1986): 156-66.

Barr, R., and R. Dreeben. How Schools Work. Chicago: University of Chicago Press, 1983. Bates, P. "Desegregation: Can We Get There from Here." Phi Delta Kappan 72 (1990): 8- 17. Bidwell, C. E. "The School as a Formal Organization." In Handbook oforganiza- tions, edited by J. G. March. Chicago: Rand McNally, 1965.

Bryk, A. S., and V. E. Lee. "Is Politics the Problem and Markets the Answer? An Essay Review of Politics, Markets, and America's Schools." Economics of Education Review 11 (1992): 439-51.

Bryk, A. S., V. E. Lee, and P. B. Holland. Catholic Schook and the Common Good. Cambridge, Mass.: Harvard University Press, 1993. Bryk, A. S., and S. W. Raudenbush. Hierarchical Linear Models: Application and Data Analysir Methods. Newbury Park, Calif.: Sage, 1992. Bryk, A. S., and Y. M. Thum. "The Effects of High School Organization on Dropping Out: An Exploratory Investigation." American Educational Research Journal 23 (1989): 353-83.

Carnegie Council on Adolescent Development. Turning Points: Preparing Amer- ican Youth ,for the 21st Centuq: The Report on the Tmk Force on Education of Young Adolescents. New York: Carnegie Corporation of New York, 1989.

Chall, J. S., V. A. Jacobs, and L. E. Baldwin. The Reading Crisis: Why Poor
Children Fall Behind. Cambridge, Mass.: Harvard University Press, 1990.
Coleman, J. S. "The Relations between School and Social Structure." In The
Social Organization of Schools: New Conceptualizations of the Learning Process,

edited by M. T. Hallinan. New York: Plenum, 1987. Coleman, J. S. "Social Capital in the Creation of Human Capital." American Journal of Sociology, 94, suppl. (1988): S95-S110. Coleman, J. S., E. Q. Campbell, C. J. Hobson, J. McPartland, A. M. Mood,

F. D. Weinfeld, and R. L. York. Equality ofEducational Opportunity. Washington, D.C.: Government Printing Office, 1966.

Comer,J. S. School Power: Implication of an Intervention Project. New York: Free Press, 1980.

326 American Journal of Education

Comer, J. S. "Educating Poor Minority Children." Scient$c American 259 (1988): 42-48. Committee for Economic Development. Children in Need: Investment Strategies for the Educationally Disadvantaged. New York: Committee for Economic De- velopment, 1987. Cusick, P. A. The Egalitarian Ideal and the American High School. New York: Longman, 1983. Department of Health and Human Services. "Annual Update of Poverty In- come Guidelines." Federal Register 52 (1 987): 5340-4 1.

Dole, J. A., G. G. Duffy, L. R. Roehler, and P. D. Pearson. "Moving from the Old to the New: Research on Reading Comprehension Instruction." Review of Educational Research 61 (199 1): 239-64.

Eccles, J. S., S. Lord, and C. Midgley. "What Are We Doing to Early Adoles- cents? The Impact of Educational Contexts on Early Adolescents." American Journal of Education 99 (1991): 521 -42.

Epstein, J. L. "Home and School Connections in Schools for the Future: Implications of Research on Parent Involvement." Peabody Journal ofEduca- tion 78 (1985): 373-80.

Epstein, J. L. "Parents' Reactions to Teacher Practices of Parent Involvement." Elementary School Journal 86 (1986): 277-94.

Epstein,J.L. "Effects on Student Achievement of Teachers' Practices of Parent Involvement." In Advances in ReadinglLanguage Research. Vol. 5, Literacy through Family, Community, and School Interaction, edited by S. B. Silvern and

B. A. Hutson. Greenwich, Conn.: JAI, 1991.

Epstein, J. L., J. M. McParland, and D. J. MacIver. Hopkins Enhancement Suruq of NELS:88 Middle Grades Practices: Codebook and Data Collection Instruments. Baltimore: Johns Hopkins University Center for Research on Effective Schooling for Disadvantaged Students, 199 1.

Garnoran, A. (1991). "Schooling and Achievement: Additive versus Interactive Models." In Schools, Classrooms, and Pupils. International Studies of Schooling from a Multilevel Perspective, edited by S. W. Raudenbush and J. D. Willms.

San Diego, Calif.: Academic Press. Grant, C. A., and C. E. Sleeter. "Race, Class, and Gender and Abandoned Dreams." Teachers College Record 90 (1988): 19-40.

Haller, E. P., D. A. Child, and H. J. Walberg. "Can Comprehension Be Taught? A Quantitative Synthesis of "Metacognitive" Studies." Educational Researcher 17 (1988): 5-8.

Hallinger, P., and J. F. Murphy. "The Social Context of Effective Schools." American Journal of Education 94 (1986): 328-55.

Hull, G. A. "Research on Writing: Building a Cognitive and Social Understand- ing of Composing." In Toward a Thinking Curriculum: Current Cognitive Re- search, edited by L. B. Resnick and L. E. Klopfer. Washington, D.C.: Ameri- can Society for Curriculum Development, 1989.

Ingels, S. J., S. Y. Abraham, K. Rasinski, R. Karr, B. D. Spencer, and M. R. Frankel. National Education Longitudinal Study of 1988: Data File User's Man- ual: Base Year: Student Component. National Center for Education Statistics report no. 90-464. Washington, D.C.: U.S. Department of Education, Office of Educational Research and Improvement, 1989.

Knapp, M. S., and M. Needels. "Review of Research on Curriculum and Instruction in Literacy." In Better Schooling for the Children of Poverty: Alternatives to Conventional Wisdom. Vol. 2, Commissioned Papers and Litera- ture Review, edited by M. S. Knapp and P. M. Shields. Washington, D.C.:

May 1994 327

U.S. Department of Education, Office of Planning, Budget, and Evalua- tion, 1990.

Knapp, M. S., and B. J. Turnbull. Better Schoolingfor the Children of Poverty: Alternatives to Conventional Wisdom. Vol. 1, Summav. Washington, D.C.: U.S. Department of Education, Office of Planning, Budget, and Evaluation, 1990.

Kozol, J. Savage Inequalities: Children in America's Schools. New York: Crown, 1991. Lareau, A. Home Advantage: Social Class and Parental Intervention in Elementa~ Education. Philadelphia: Falmer, 1989. Lee, V. E., and A. S. Bryk. "A Multilevel Model of the Social Distribution of High School Achievement." Sociology ofEducation 64 (1989): 78-94.

Lee, V. E., A. S. Bryk, and J. B. Smith. "The Organization of Effective Second- ary Schools." In Review ofResearch in Education, vol. 19, edited by L. Darling Hammond. Washington, D.C.: American Education Research Association, 1993.

Lee, V. E., and J. B. Smith. "Effects of School Restructuring on the Achieve- ment and Engagement of Middle-Grade Students." Sociology ofEducation 66 (1993): 164-87.

Mahard, R. E., and R. L. Crain. "Research on Minority Achievement in Deseg- regated Schools." In The Consequences of School Deseg-regation, edited by

C. H. Rossell and W. D. Hawley. Philadelphia: Temple University Press, 1983.

Moles, 0.C. "Who Wants Parent Involvement? Interest, Skills, and Opportu- nities among Parents and Educators." Education and Urban Society 19 (1987): 137-45.

Natriello, G., E. L. McDill, and A. M. Pallas. Schooling Disadvantaged Children: Racing Against Catastrophe. New York: Teachers College Press, 1990. Newmann, F. M. "Reducing Student Alienation in High Schools: Implications of a Theory." Harvard Educational Review 51 (1981): 546-64. Newmann, F. M. "Linking Restructuring to Authentic Student Achievement." Phi Delta Kappan 72 (1991): 458-63. Newmann, F. M., ed. Student Engagement and Achievement in American Secondaq Schools. New York: Teachers College Press, 1992. Newmann, F. M. "Beyond Common Sense in Educational Restructuring: The Issues of Content and Linkage." Edztcational Researcher 22 (1993): 4- 13.

Oakes, J. Keeping Track. New Haven, Conn.: Yale University Press, 1985.

Ogbu, J. ti. "Minority Status and Literacy." In Literacy: An Ozieruiew of 14 Experts, edited by S. K. Graubard. New York: Hill & Wang, 1991.

Orland, M. E. "Demographics of Disadvantage: Intensity of Childhood Poverty and Its Kelationship to Educational Achievement." In Access to Knowledge: An Agendafor Our Notion's School., edited by J. I. Goodlad and P. Keating. New York: College Entrance Examination Board, 1990.

Resnick, L. B. "Literacy in School and Out." In Literacy: An Ozierz~ieu~ of 14 Experts, edited by S. R. Graubard. New York: Hill & Wang, 1991. Rosenthal, R., and R. L. Rosnow. Essentials ofBehauiora1 Reseorch: Methoh and Data Analjsis. New York: McGraw-Hill, 1984. Scott-Jones, D. "Family Influences on Cognitive Development and School Achievement." Kez~iew ofResearch in Education 11 (1984): 259-306.

Slavin, R. E. "Ability Grouping and Student Achievement in Elementary Schools: A Best-Evidence Synthesis." Reuirw of Educational Kesrarch 57 (1987): 293-.'i36.

328 American Journal of Education

Slavin, R. E., N. A. Madden, N. L. Karweit, B. J. Livermon, and L. Dolan. "Success for All: First-Year Outcomes of a Comprehensive Plan for Re- forming Urban Education." American Educational Research Journal 27 (1990): 255-78.

Snow, C. E., W. S. Barnes, J. Chandler, I. F. Goodman, and L. Hemphill. Unfuljilled Expectations: Home and School Influences on Literacy. Cambridge, Mass.: Harvard University Press, 199 1.

Stevenson, D. L., and D. P. Baker. "The Family-School Relation and the Child's School Performance." Child Developnent 58 (1987): 1348-57.

Stevenson, H. W., C. Chen, and D. Uttal. "Beliefs and Achievement: A Study of
Black, White, and Hispanic Children." Child Development 61 (1990): 508-23.
Vergon, C. B. "School Desegregation: Lessons from Three Decades of Experi-

ence." Education and Urban Society 23 (1990): 22-49. Wilson, J. W. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: University of Chicago Press, 1987.

May 1994 329

  • Recommend Us