English Language Learner Classmates and the Classroom Social Skills of Students with Disabilities
by Michael A. Gottfried & Morgan S. Polikoff - 2015
Background/Context: Though the development of social skills in kindergarten is critical, a research gap exists in how the context of the general education classroom may influence the social skills outcomes of students with disabilities: None have considered the role of peer effects in this domain. This gap is critical to address, as multiple high-needs groups are increasingly present in the same general education classroom settings.
Purpose/Objective: This study asks two key research questions: (1) In kindergarten, to what extent do the classroom social skills outcomes of children with disabilities differ based on the number of ELL classmates? (2) In kindergarten, to what extent do the classroom social skills outcomes of ELL students differ based on the number of classmates with disabilities?
Population/Participants: The data are sourced from the Early Childhood Longitudinal Study-Kindergarten Class (ECLS-K), which is a nationally representative sample of students, teachers, and schools. Information was first collected from kindergartners (as well as parents, teachers, and school administrators) from U.S. kindergarten programs. This study utilizes data collected at the fall and spring of kindergarten.
Research Design: This study combines secondary data analyses and quasi-experimental methods. There are three social skills outcomes: (1) approaches to learning, (2) interpersonal skills, and (3) self control. The study begins with a baseline, linear regression model. To address issues pertaining to omitted variable bias, the study employs multilevel fixed effects modeling.
Findings: The coefficients indicate that students with disabilities tend to have improved social skills with an increase in the number of ELL classmates. The effects remain significant even after accounting for multiple omitted variable biases. Notably, the reverse relationship does not hold: The number of classmates with disabilities has no significant influence on the outcomes of ELL students.
Conclusions/Recommendations: This research offers more in-depth insight into how the classroom context and the effects of classmates may have a unique relationship for specific high-needs groups such as students with disabilities—a strand of research in this area that is often overlooked. School practices can thus be guided by determining not simply if one group of students performs better or worse on average, but rather by asking, better or worse for whom in particular?
Research has repeatedly documented that, on average, students with disabilities have lower academic and behavioral outcomes compared to their non-disabled counterparts. For instance, in a nationally representative dataset, approximately 66% of students with disabilities in eighth grade scored below basic on the 2005 National Assessment of Educational Progress in reading and math, as compared to approximately 25% of non-disabled students (U.S. Department of Education, 2007). It has been long established that students with disabilities are also more likely to drop out of school (Blackorby & Wagner, 1996). Research shows that students with disabilities are, on average, lagging behind their non-disabled counterparts on multiple measures of socioemotional attainment (Phelps & Hanley-Maxwell, 1997), including life satisfaction (Blackorby & Wagner, 1996).
In an attempt to support the positive academic growth and development for students with disabilities, special education has been a focus for state and federal education policy. As one major strategy since the passage of IDEA, federal policy has increasingly encouraged the placement of students with disabilities into the general education classroom with their non-disabled peers. Thus, as the proportion of students identified as disabled continues to grow (with an approximate 17.1% average growth in developmental disabilities for instance; Boyle et al., 2011), there will be an increasingly larger presence of students with disabilities across the nation who receive some or all of their instruction from within the general education classroom (U.S. Department of Education, 2012). That is to say, classrooms that previously did not contain students with disabilities may now be experiencing compositional changes; therefore, even an increase by one or two students may represent a large change for many classrooms.
Given these trends, we might expect to see a continued presence of students with disabilities in general education classrooms across the United States in the future. And while there is some evidence that general education classrooms are, on average, raising achievement for students with disabilities (Baker, Wang, & Walberg, 1994; Carlberg & Kavale, 1980; Lindsay, 2007; Thurlow, Quenemoen, & Lazarus, 2012), there is little exploration of how these classrooms might be influencing other outcomes (Lindsay, 2007) as discussed below.
Simultaneously, a second trend is occurring in our nations schools: an increasing presence of students in the United States whose primary language is not English. Indeed, the growth in the number of English Language Learner (ELL) students has surpassed the growth in the number of non-ELL students (Fry, 2008). According to the National Center for English Language Acquisition (NCELA, 2010), the national population of ELL students has increased more than 53% between 1997 and 2007. In contrast, the overall school population has only grown by 8.5% over that same period. With this rate of growth, ELL students are projected to account for at least 25% of the schooling population by 2015 (National Education Association, 2008). Importantly, while ELL students traditionally were geographically located in a few states (e.g., California), the trend in U.S. immigration is no longer regionalized. As is the case for students with disabilities, ELL students are now present in classrooms across the United States where they previously were not found. Thus, schools and classrooms that did not historically have ELL students are now experiencing demographic changes (e.g., Allentown, Pennslyvania; Fresno, California; and Lowell, Massachusetts).
Given this spread in immigration patterns across the United States, state and federal educational agencies over recent decades have also been considering how best to meet the needs of ELL students. While many states continue to have separate bilingual education (e.g., Texas and Wyoming), many other states have moved away from bilingual education toward an English-only approach that can be segregated (e.g., Arizona) or not (e.g., Massachusetts). There has also been a push from the federal level for mainstreaming of ELL students. Federal policies, including No Child Left Behind (NCLB), have mandated or induced districts and schools to educate ELL students with their non-ELL peers to the maximum extent possible in general education classrooms. Hence, in conjunction with special education inclusion, there has been tremendous growth in ELL presence in K12 classrooms across the nation as a whole.
In recent years, many educational stakeholders question whether the context of an increasingly diverse and mainstreamed general education classroom can adequately provide services to so many high-needs students at once (Moon, Todd, Morton, & Ivey, 2012; Rule, Stefanich, Haselhuhn, & Peiffer, 2009; Supalo et al., 2008). For instance, concerns are often raised that teachers are underprepared to educate students with disabilities (Moon et al., 2012) and are also underprepared to educate ELL students (Reeves, 2006). Thus, mainstreaming practices and changes in classroom diversity may exacerbate the strain on classroom resources as teachers must deliver content to an increasingly wider range of students (Tichenor, Heins, & Piechura-Couture, 2000). Indeed, many are concerned that it is becoming increasingly difficult to balance the needs of so many students in general education classrooms, particularly those students with disabilities (De Cohen, Deterding, & Clewell, 2005; Hayworth, 2009; Supalo et al., 2008; Wiley & Wright, 2004).
As research upholds that the skills that children acquire during kindergarten are highly predictive of future outcomes (Alexander, Entwisle, & Dauber, 1993; Juel, 1988; Pianta & Walsh, 1996; Smith, 1997), stakeholders are concerned about these issues pertaining to the effects of the general education classroom context as early as school entry (i.e., kindergarten). Research has predominantly focused on how features of the kindergarten classroom context may influence academic achievement (see, e.g., Gullo, 2000; Karweit, 1992; Lee et al., 2006; NEA, 2008; Walston & West, 2004; Weast, 2004); less overall is known about the effects of the kindergarten classroom setting on social skill development. That said, social skills are especially critical to foster in kindergarten. Kindergarten has been documented as a critical developmental time in education when it comes to social development (Olson et al., 2005; Posner & Rothbart, 2000). Indeed, research supports that gaining proper social functioning in kindergarten may set the trajectory for longer-term outcomes (Juel, 1988; Pianta & Walsh, 1996; Smith, 1997). Thus, determining which contextual factors in kindergarten might be related to fostering these social skills may have significant implications for students across their lifespans.
Though the development of social skills in kindergarten is critical, prior to this study, a research gap existed in how the context of the general education classroom may influence the social skills outcomes of students with disabilities: None have considered the role of peer effects in this domain. This gap is critical to address, as multiple high-needs groups are increasingly present in the same general education classroom settings, as described above. Thus, knowing how peer context plays a role in influencing outcomes will continue to become increasingly influential to ensure the proper social functioning of all students. This may be particularly useful to address for students with disabilities, as some subgroups of disabilities may have challenges in the general education classroom environment (e.g., students with emotional or behavioral disorders) (Fletcher, 2010; Hazel & Schumaker, 1988; Kavale & Forness, 1995; Kavale & Mostert, 2004). As such, our inquiry begins with:
Research Question 1: In kindergarten, to what extent do the classroom social skills outcomes of children with disabilities differ based on the number of ELL classmates?
To date, there has been very little research examining the effect of having high-needs classmatesof this research, all work has focused on the outcomes of the typical general education student (see, e.g., Cho, 2012; Fletcher, 2010; Hanushek, Kain, & Rivkin, 2002). Critically, no empirical study has examined the relationship between the presence of a high-needs student subgroup in the general education classroom and the outcomes of another high-needs student subgroup in that same classroom. Of the research that has been conducted on the peer effects of having high-needs classmates, the results have been mixed. Gottfried (2014a) finds that for the typical student, having classmates with disabilities reduces socio-emotional outcomes in early elementary school classrooms. Fletcher (2010) finds an analogous negative peer effect on achievement from having classmates with emotional and behavioral disorders. On the other hand, Hanushek et al. (2003) find positive achievement effects of classmates with special needs on non-disabled students. As for the peer effects of having an ELL classmate, only two studies exist: Cho (2012) found a negative effect on achievement for typical non-ELL students. On the other hand, Gottfried (2014b) found a positive effect of having ELL classmates on the social skills of typical non-ELL students.
As mentioned, no research has considered how the presence of one high-needs group might associate with another high-needs groups outcomes. That said, there is some limited evidence that high-needs students do affect the outcomes of typical students, as mentioned above. Relying on this scant body of research that does find statistically significant effects of having high-needs classmates, we hypothesize that ELL students would influence the outcomes of students with disabilities. Whether this effect is positive or negative is the focus of this study.
From the positive perspective, there are two potential mechanisms through which ELL classmates may improve the classroom social skills of students with disabilities: peer interactions and classroom resources. As for peer interactions, having ELL classmates can help students with disabilities to interact with other groups of high-needs students like ELL students. Here, a similar mechanism exists when considering how the presence of ELL students affects general education students (see, e.g., Gottfried, 2014b). In the present study, the presence of ELL classmates would effectively allow for students with disabilities to increase their levels of tolerance and patience and to broaden their understanding and patience with individual differences (Cho, 2012; Williams & Downing, 1998)much like it might for students without disabilities (Gottfried, 2014a). Thus, each student may nonetheless experience a boost in social skills (from whatever initial developmental level they find themselves) through the experience of working alongside diverse students, such as ELL students. As for classroom resources, with the inclusion of two high-needs groups in the same classroom, there may be additional supports and services (Hanushek et al., 2002; Lipsky & Gartner, 1995) such as multiple teacher aides to address multiple demands from multiple types of high needs (Winters & Greene, 2007). These additional supports and resources thus encourage a reshuffling of the teachers time allocation, thereby allowing the teacher to sufficiently address the needs of all students in the classroom. This is critical for development, as greater teacher-to-individual-student interactions have been supported in the research as nurturing greater levels of classroom social skills (Kontos & Wilcox-Herzog, 1997).
However, the influence of ELL classmates on students with disabilities might play out in the negative direction, again both through peer interactions and classroom resources. As for peer interactions, prior research has found that an increase in diversity might fracture a classroom environment (Banks & Banks, 1995). Thus, one potential negative mechanism might arise from being in classrooms with multiple groups of high-needs students, e.g., ELL classmates as well as students with disabilities in the same classroom; namely, more diversity may exacerbate a classroom fracture, thereby potentially deteriorating social skills for all students in the classroom. As for classroom resources, as mentioned above, another potential negative mechanism might arise from teaching resources being spread thin with the inclusion of multiple high-needs groups in the same classroom (Hayworth, 2009; Tichenor et al., 2000): Having to address the needs of ELL classmates as well as the needs of students with disabilities, teachers may not be able to adequately address all unique needs. Consequently, it might be the case that classrooms are not appropriately structured to provide all students with an equitable amount of teaching time and attention, thereby increasing classroom disengagement among students or straining the teachers ability to properly interact with others (Hayworth, 2009; Karabenick & Noda, 2004).
To address this primary research question, this study relies on data from the ECLS-K, a large-scale national dataset that links individual students to the characteristics of their families, classrooms, and teachers. Because of this wide range of data availability, this study will be able to utilize a methodology (described in the next section) that is supported in general educational research (Schneider et al., 2007), in classroom peer effects research (Hanushek, Kain, Markman,, & Rivkin, 2003; Zimmer & Toma, 2000), and in peer effects research that utilizes ECLS-K (Cho, 2012; Fletcher, 2010). Moreover, utilizing such detailed national data also enables multiple tests of the main findings, including the examination of additional outcomes beyond social skills, as well as a test of the robustness of the social skills ratings.
This study is put forth as an examination of the effect of classroom context (i.e., the number of ELL classmates) on the outcomes of students with disabilities. That said, policy implications cannot be properly made without knowing reciprocal effects. Hence, a second research question is put forth as follows:
Research Question 2. In kindergarten, to what extent do the classroom social skills outcomes of ELL students differ based on the number of classmates with disabilities?
As mentioned, in most states, classrooms in aggregate are trending toward having a presence of both students with disabilities and ELL students in general education classrooms. Thus, a complete assessment would examine the relationships of classmates with disabilities to the outcomes of ELL students. Knowing if a relationship exists in both directions would allow the field to begin considering how to make more confident decisions about the total distribution of students in classrooms.
In sum, this study contributes to knowledge in this area by providing a deeper understanding of how the context of peer effects in kindergarten may relate to a range of outcomes for students with high needs. This peer effects analysis is innovative because it employs individual- and multi-level national data that allows for student data to be linked to precise classroom attributes (in addition to other family, classroom, and teacher metrics). By isolating how the classroom context relates to these outcomes and quantifying these associations with more methodological rigor and precision than previously used in the field, the findings yield insight into the underlying predictors of the outcomes of students with disabilities in early schooling experiences. This will enable policymakers and practitioners to more efficiently identify and mitigate individual and classroom contextual risk factors in order to consider how to design and foster supportive educational environments from the very start of schooling.
Data in this study were sourced from the Early Childhood Longitudinal Study-Kindergarten Class (ECLS-K). This survey was developed by the National Center for Educational Statistics (NCES) and is a national sample of students, families, teachers, classrooms, and schools. The ECLS-K used a three-stage stratified sampling design, in which geographic region represented the first sampling unit, public and private schools represented the second sampling unit, and students stratified by race/ethnicity represented the third sampling unit. Hence, the children in ECLS-K come from a diversity of school types, socioeconomic levels, and racial and ethnic backgrounds. Germane to this current study, NCES initially collected data for kindergartners by surveying parents, teachers, and school administrators from approximately 1,000 kindergarten programs in the fall of kindergarten.
To detect if there are differences in social skills based on varying numbers of ELL classmates, the analyses focus on social skills outcomes measured in the spring of kindergarten. Data that described both concurrent and past factors were sourced from fall and spring kindergarten teacher and parent surveys and were used as independent variables, as described below.
There were approximately N = 9,630 student observations available for the analyses in this study. The analyses in this study were limited to first-time kindergartners and children who had non-missing information on social skills scales. Comparing children in the analytic sample versus those excluded did not generate any significant differences. Note that all sample sizes presented in this study were rounded to the nearest 10 due to NCES regulations.
Outcome Measures: Classroom Social Skills
The means and standard deviations of both dependent and independent variables utilized in this study are presented in Table 1. This study relies on utilizing social skills scales created by NCES for the purposes of evaluating students in the ECLS-K dataset. NCES did so by modifying a version of the Social Skills Rating System (SSRS; Greshan & Elliott, 1990) to measure a childs behavior or social skillset. These original SSRS scales have been considered to be the most comprehensive social skill assessment that can be widely administered in large surveys (Demaray et al., 1995). For the purposes of ECLS-K data analysis, NCES modified the original scales and created its own Teacher Social Rating Scale (SRS). Meisels, Atkins-Burnett, and Nicholson (1996) provide detail on these modifications from SSRS to the ECLS-K SRS. That said, it is not possible to obtain individual items in each of the scales described hereeven with the restricted-users access to the data, only aggregate, final scales are provided to the researcher.
Three teacher-rated SRS classroom social skills scales were utilized in this study. Note that teacher-rated scales were selected purposefully. In survey research, having teachers report on students socio-emotional outcomes is supported as the most relevant source for students school functioning compared to relying on other sources for this information, such as parents (Konold & Pianta, 2007; Lee et al., 2013; Waterman, McDermott, Fantuzzo, & Gadsden, 2012).
NCES provided three classroom social skills scales in the ECLS-K dataset: (1) approaches to learning, (2) interpersonal skills, and (3) self control. The scale on approaches to learning rates a childs frequency of organization, eagerness to learn new things, independent work ability, adaptability to change, persistence in completing tasks, and ability to pay attention. The interpersonal skills scale measures the frequency with which a child has been getting along with people, forming and maintaining friendships, helping other children, showing sensitivity to the feelings of others, and expressing feelings, ideas, and opinions in positive ways. The scale on self control measures the frequency of the students ability to control his or her temper, respect for others property, acceptance of peer ideas, and handling peer pressure.
NCES developed each construct as a continuous measure: Each Likert scale is based on an average of a series of individual questions, where replies to each question ranged from 1 (never) to 4 (very often), thereby making a score of 4 the most favorable outcome. NCES reports that these scales have high construct validity as assessed by testretest reliability, internal consistency, inter-rater reliability, and correlations with more advanced behavioral constructs. While NCES does not report all of these metrics in its manual (even at the restricted-user level), it does report that the split half reliability for approaches to learning is 0.89 in both fall and spring assessments, for interpersonal skills is 0.89 in both fall and spring assessments, and for self control is 0.79 in fall and 0.80 in spring.
Table 2 presents descriptive statistics on the outcome measures for each group of interest in this study: students with disabilities, students without disabilities, ELL students, and native English speakers. The table demonstrates that there are statistical differences between each respective group, which motivates the following analysis. Additionally, note that students with disabilities have the lowest social scales means of all groups. This is consistent with previously mentioned findings that would suggest that social skills are significant issues for some students with disabilities (Hazel & Schumaker, 1988; Kavale & Forness, 1995; Kavale & Mostert, 2004).
Disability status. Students with disabilities were identified by their school records on file. In the ECLS-K study, a student was identified as having a disability if the student had an Individualized Education Plan (IEP). Note that this includes both learning and non-learning (physical) disabilities. This equates to approximately 10% of the students in the sample, as indicated in Table 1. Moreover, this approximates the number of students with disabilities in the U.S. population, in general (NCES, 2011). Hence, this provides confidence in this dataset in that it is reflecting the demographic makeup of U.S. students.
Number of ELL classmates. The key independent variable in this analysis is the total number of ELL classmates. This was sourced from the spring teacher survey, in which teachers reported the total number of students with limited English proficiency in their classrooms. To be specific, in the survey, teachers were asked to report on the following: How many children with limited English proficiency (LEP) do you have in each of your classes? The response space allowed for teachers to write in the exact number of students: The responses ranged from 0 to 32. Note that all classroom peer information was sourced from the teacher. To be clear, the teacher reported the total number of students with limited English proficiency in the entire classroom, not simply those in the ECLS-K study.
The strength of utilizing a dataset like ECLS-K is the ability to employ an extremely wide range of control variables that may explain the variance in social skills outcomes.
Student demographics and home data. At the level of the student, the set of control variables included a commonly accepted set of demographic variables, including: gender, race, age, and individual-level ELL status. Additionally, the student level included a measure of a students health, given that health is supported as being correlated with socio-emotional and psychological outcomes (Drotar, 1997). As such, this study included a parental rating of the childs physical health upon kindergarten entry, either excellent, very good, good, fair, or poor.
Additionally, because ELCS-K includes parent surveys, this study can incorporate data on students school-related home lives. Consistent with prior research examining social skills outcomes in kindergarten (e.g., Loeb et al., 2007; Rhum, 2004), additional measures included the frequency of reading books, the total number of books, and whether or not the child attended out-of-home prekindergarten care.
Family data. The ECLS-K parent survey provides a great opportunity to fold key household measures into this study. Socioeconomic status (SES), including multiple measures of SES, is critical to the study because there is evidence that SES is highly correlated with socio-behavioral outcomes (Aber, Bennett, Conley, & Li, 1997; Brooks-Gunn & Duncan, 1997). In this study, commonly accepted empirical measures of SES were employed as control variables (e.g., Cho, 2012; Loeb et al., 2007; Rhum, 2004), including a five-scale SES composite created by NCES, the mothers education, and an indicator for whether or not the family was at or below the poverty threshold. Additional measures, as presented in Table 1, include the number of adults and the number of siblings in the household. Finally, parental involvement in both the home and schooling lives of the child was captured with six measures. Of these, four were four-point scales determining the frequency with which the responding parent sang to the child, told stories to the child, engaged the child to do chores in the house, and played games with the child. The final two parental measures were binary, indicating whether the parent used spanking as punishment and whether the parent lived in the current home due to the childs school location.
Classroom and teacher data. Given that students are assigned one classroom and one teacher in kindergarten, it is possible to identify precise classroom characteristics for each student. Prior research has indicated that several classroom characteristics serve as critical factors in student outcomes (particularly socio-emotional outcomes) and are thereby employed as control variables in this study. They include class size (see, especially, Dee & West, 2012), gender distribution of the classroom (Hoxby, 2000), average classroom reading test scores (measured at kindergarten entry), and the number of classmates with disabilities (Gottfried, 2014a). As consistent with prior empirical work using ECLS-K (Fletcher, 2010), a measure of classroom racial demographics is also included as a demographic classroom control variable.
As for teacher control variables, prior research finds that teacher characteristics correlate with young students social skills development (Coplan & Prakash, 2003; Kontos & Wilcox-Herzog, 1997). Based on prior empirical research using ELCS-K, commonly accepted teacher characteristics employed as control variables include teacher race, gender, and years of experience (Cho, 2012; Fletcher, 2010; Neidell & Waldfogel, 2010). However, given that this study investigates the relationship between students with disabilities and ELL classmates, additional teacher characteristics are included, such as the total number of course units the teacher has taken in special education as well as the total number of course units in ESL.
The first column of Table 3 presents two series of partial correlation coefficients. The first column presents the partial correlation coefficients and their significance values between having a disability and the other independent covariates in Table 1. This table is included in order to determine if students with disabilities are more or less likely to embody certain basic demographic characteristics or have come from families that might bias the data in any particular direction. The second column presents the partial correlation coefficients and their significance values between the number of ELL classmates and the other independent variables in Table 1.
Of great importance is the small correlation value between being a student with a disability and the number of ELL classmates. Had this correlation value been high, it may have been hypothesized that school administrators might sort all students with high needs into the same classroom. However, the correlation coefficient in the first row of the table is -0.01, indicating that there is nothing systematic in the data between having a disability and having greater or fewer ELL classmates.
The second column of Table 3 presents the correlation coefficients between the number of ELL classmates and the other independent covariates. As consistent with the first column, there are only weak-to-zero correlation values throughout the second column. Hence, classrooms with higher numbers of ELL classmates do not appear to be systematically related in any meaningful way to other observable characteristics. In other words, there does not appear to be a sorting of ELL students in a way that might influence the estimates to follow.
Though almost zero in numerical value, the partial correlation coefficient presented in Table 3 between having a disability and the number of ELL classmates merits further exploration. Table 4 examines the possibility of the distribution of students in classrooms across two dimensions: students with disabilities and the number of ELL classmates. In Table 4, each column is a separate regression model with the dependent variable labeled at the top. Nine dependent variables are presented: All were measured and sourced from the kindergarten entry survey wave. They represent characteristics and traits that school administrators might use to non-randomly assign students at the start of the kindergarten school year. The independent variables are all other covariates listed from Table 1. Included are binary indicators for each school, which as explained in more detail below, enables the model to strictly look at within-school variation in classroom composition.
The table is divided into two panelsthe top panel assesses whether there is sorting by ELL classmates for the sample of students with disabilities, and the bottom panel does so for all other students in the sample. With these nine dependent variables, if there was evidence of sorting of students by the number of ELL classmates, then the number of ELL classmates would be statistically significant in relation to the dependent variables measured at kindergarten entry (Neidell & Waldfogel, 2010). However, both panels show that this is not the case. Across all columns in this table, there is no evidence that nine characteristics of any student at school entrywith or without a disabilityis related to number of ELL classmates in kindergarten. This null finding of no systematic assignment in kindergarten is consistent with prior peer research using ECLS-K data (e.g., Aizer, 2008; Neidell & Waldfogel, 2010). This finding thus supports the empirical strategy described in the following section.
Note that in an ancillary test, the variable number of ELL classmates is replaced with a binary indicator for having any ELL classmates at all. However, the results remain consistent with those presented in Table 4. Thus, neither students with disabilities nor students without disabilities are systematically placed in classrooms with ELL students. The findings in Table 4 also present evidence that there is no sorting of students by disabilities. This finding supports the analytic approach described in the proceeding section.
Baseline model. Estimating whether there is a relationship between having ELL classmates and the classroom social skills for students with disabilities begins with a baseline linear regression model, presented as follows:
SSijk = β0 + β1Dijk*ELLijk+ β2Sijk + β3Fjk + β4Cjk + εijk (1)
where SS is one of three classroom social skills SRS scales for student i in kindergarten classroom j in school k.
The key coefficient in this study is β1, which represents the association of the number of ELL classmates for students with disabilities. Technically, β1 is the interaction between an indicator for having a disability and the number of ELL classmates. Deriving β1 as an interaction between Dijk (having a disability) and ELLijk (having ELL classmates) is grounded in the empirical literature as an appropriate estimation technique for assessing heterogeneity in classroom or schooling predictors as differentiated by a specific individual student-level characteristic, which in this case is having a disability (e.g., Yamauchi & Leigh, 2011).
The term Sijk includes the main predictors Dijk as well as ELLijk. It also includes all student information presented in Table 1 as well as a one-survey wave lagged measure of the outcome (i.e., the outcome measured at kindergarten entry). All family data, including SES, is incorporated in the term Fjk. Classroom and teacher measures are included in Cjk. The error term ε includes all unobserved determinants of the outcome. Empirically, this latter component is estimated with robust standard errors, adjusted for classroom clustering. Because students are nested in schools by classroom and hence share common but unobservable characteristics and experiences, clustering student data at the classroom level provides for a corrected error term given this non-independence of individual-level observations and given that the treatment of having ELL classmates occurs at the level of the classroom.
The baseline model, as presented in equation (1), included a wide array of independent covariates, including student, family, and classroom and teacher factors. Having access to these observable measures in the ECLS-K dataset will certainly aid in the reduction of any biases in estimating β1. That said, however, there may be unobserved school-level factors, processes, and policies that may be influencing the estimate of β1.
In one hypothetical example, students might have highly involved parents who choose to send their children to schools where there is a greater probability of interaction with diverse students, such as ELL students. The parents would make this choice based on prior research showing that students in schools with greater diversity have higher schooling outcomes (Fryer & Levitt, 2004). Simultaneously, however, these same highly involved parents might be making additional investments at home that would boost their childrens social skills. If it were common for all students in a school to have parents such as these, then the peer effect of having ELL classmates would be confounded by a high level of parental involvement. As a second example, principals at some schools may have invested in a greater number of policies and practices to boost classroom social skills. These same principals might also be more likely to have greater inclusion policies for all students (e.g., ELL and disability), hypothetically speaking. In such a case, one might overestimate any positive relationship between having ELL classmates and students social skills outcomes.
These are two of many potential scenarios in which school-level factors could be influencing both the key independent variable as well as the classroom social skills dependent variables. Hence, a second empirical model was employed to test the robustness of baseline estimates of β1. Notably, a second model includes school-level fixed effects:
SSijk = β0 + β1Dijk*ELLijk+ β2Sijk + β3Fjk + β4Cjk + δk+ εijkt (2)
where δk indicates the use of school fixed effects for each school k. Technically speaking, the term δk is a set of binary variables that indicates whether a student attended a particular school (for each school variable in the dataset, 1 indicates yes, and 0 indicates no). This set of school indicator variables leaves out one school as the reference group (e.g., this process is analogous to creating indicator variables for race, where one racial category is left out as the reference group).
The use of school fixed effects is compelling in this study and is supported as an appropriate technique in prior empirical educational research pertaining to examining classmate effects both with the ECLS-K data as well as with other large-scale datasets (e.g., Burke & Sass, 2008; Cho, 2012; Fletcher, 2010; Hanushek et al., 2003). The importance of school fixed effects δk is that they control for all unobserved school-level influences because they hold constant omitted school-specific factors, such as aggregate parental involvement and administrators’ inclusion policies. In doing so, the key source of variation used to identify the relationship between ELL classmates and students social skills occurs across classrooms within each school. The use of school fixed effects, as presented in this section, has been supported in the literature. Consistent with prior research (Aizer, 2008; Neidell & Waldfogel, 2010) and with Table 3, this study also upholds that there is little evidence of within-school sorting in the ELCS-K dataset. Thus, this method is most appropriate when evaluating the effects of classmates on social skills outcomes (see, e.g., Neidell & Waldfogel, 2010).
In an ancillary test, equation (2) is supplemented with U.S. region fixed effects as well as school-by-U.S. region fixed effects. For instance, it may be the case that there are specific unobserved contextual factors in a given region of the country or in a given school and in a given region (e.g., the South) that would be influencing both the number of ELL classmates and classroom social skills outcomes. This may be particularly so in this study given historical U.S. immigration patterns and given the focus of the effects of ELL classmates. The estimates of β1, however, are extremely similar to those found using equation (2). Hence, these results are not presented, though they are available upon request.
Table 5 presents unstandardized coefficient estimates and standard errors for the specifications examining the relationship between having ELL classmates and the outcomes for students with disabilities (e.g., the three ECLS-K SRS social skills outcomes as described previously). For each outcome, both baseline and school fixed effects models are presented. Respectively, these models are based on equation (1) and equation (2). Recall that the key coefficients of interest are found in β1 (i.e., the interaction between the number of ELL classmates and being a student with a disability). These interactions are located in the first row of results in Table 5. The two rows that immediately follow the interactions present the main terms of being a student with a disability and of the number of ELL classmates. Following this main set of results are the coefficient estimates and standard errors for all other independent covariates in the model.
To begin, there are several key points. First, Table 5 suggests the baseline models, which contain only observable characteristics and a pre-test measure, can explain approximately between 40% and 50% of the outcomes. Second, the inclusion of school fixed effects improves the explained portion of the variance under all outcomes, as denoted by the upward change in the R2 values at the bottom of the table. Additionally, the Likelihood Ratio test supports the implementation of fixed effects over strictly using the baseline model. Regardless, the key interaction is statistically significant in all columns of the table, and hence, the inclusion of more complex modeling techniques does not veer from supporting this studys key hypothesis: that there is a statistically significant relationship between the number of ELL classmates and classroom social skills outcomes for students with disabilities.
Examining this relationship in more depth reveals the following interpretation for students with disabilities. The first row of results portrays a unique story for students with disabilities in terms of their relationship to ELL classmates. Hence, delineating the results by disability status in an interaction proves to be critical. The coefficient on the interaction term suggests that students with disabilities in a classroom with a greater number of ELL students tend to have higher classroom social skills compared to those students with disabilities who have fewer ELL classmates. This is depicted by consistent, statistically significant, and positive coefficients on the interaction between students with disabilities and the number of ELL classmates, which indicates that higher frequencies of approaches to learning, interpersonal skills, and self control arise for students with disabilities who have higher numbers of ELL classmates. So, while there is a negative direct relationship between having a disability and classroom social skills outcomes (as indicated in the second row of results) for those students with disabilities in classes with ELL students, it is possible to reduce this relationship based on the peer classroom context.
The effect sizes of the interaction terms in the first row are 0.02σ to 0.03σ. However, given that the range of the number of ELL students within a classroom goes up to 32 (with an average of about 1.55), presenting the interaction as an effect size at the mean only provides information about the improvement of scores given a classroom average of 1.55 ELL students. Thus, with this range in mind, a more interesting portrayal of the importance of the effect size occurs when moving beyond averages. For instance, if one looks at the interaction in the case of the largest possible number of ELL students, then the effect size changes dramatically up to 0.22σ for the approaches to learning scale. This would be practically double the magnitude of the effect size of the coefficient pertaining to having a disability. The effect size of a classroom with 16 ELL students would be essentially the same value as the effect size of having a disability in the approaches to learning model. These interpretations are fairly consistent across all models in Table 5.
Briefly examining the control variables in Table 5 provides the following interpretations. In the direction as expected, compared to girls, boys tend to exhibit lower frequencies of self control, approaches to learning, and interpersonal skills. The results across all outcomes are also delineated by additional student and family characteristics, particularly so by SES and less so by parental involvement. There is less consistency across the results for the covariates pertaining to the classroom control variables. Finally, like classroom characteristics, some teacher variables are significant, but not consistently across all teacher measures.
Note that an ancillary test examined for non-linear associations in the interaction term. When a squared term of having ELL classmates was included in the model, it was not significant, though it did not change the main, linear interaction term. Thus, only linear direct effects were presented. Also note that the models in Table 5 were tested for differences based on many of the individual-level control variables listed in the model. These results, however, did not suggest heterogeneity in the effects. This implies there is not a differential association based on gender, race, or SES that would have moderated the relationship between the number of ELL classmates and the social skills outcomes for students with disabilities. That is, the predictive relationship is comparable between categories of individual-level characteristics, like gender.
To assess the robustness of these new findings, three tests of validity are presented here. The first assesses the robustness of the scale ratings, the second examines other outcomes beyond social skills, and the third examines the role of students with disabilities on ELL students. First, as noted, the outcomes in this study are teacher-rated classroom social skills scales. Hence, there is the possibility of some degree of subjectivity in the ratings of students (DiPerna, Lei, & Reid, 2007; Galindo & Fuller, 2010). To test the robustness of the results, the sample is restricted by teacher experience. Prior research has established that a significant number of teachers leave the profession within the first few years of teaching (Ingersoll, 2002; Luekens, Lyter, & Fox, 2004). Teachers who remain teaching after these initial few years have gained more in-classroom experience. These extra years may provide them with a better foundation to accurately rate students on the social scales.
To test for the subjectivity in teacher ratings, Table 6 presents a modified version of Table 5, in which the sample is restricted to include only those students who were rated on social skills outcomes by teachers who have taught for five years or more. Five years was determined as a criterion in that Ingersoll (2002) has presented evidence that five years is a benchmark year by which time a significant portion of teachers have left the teaching profession.
The results in Table 6 are closely linked to those in Table 5. Analytically, the fixed effects models are statistically preferred to the baseline models, though regardless, all models have the power to explain between 40% and 60% of the variance in outcomes. Indeed, the coefficients of the interaction β1 are similar both in size and statistical significance. Hence, the results continue to indicate that students with disabilities tend to have higher frequencies of approaches to learning, interpersonal skills, and self control when a greater number of ELL classmates are present.
A second test examines other potential outcomes that may be related to having a greater number of ELL classmates. First, two teacher-rated problem behavioral scales are tested as outcomes. The externalizing problem behaviors scale measures the frequency with which a child argues, fights, gets angry, acts impulsively, and disturbs ongoing activities. The internalizing problem behaviors scale rated the presence of anxiety, loneliness, low self-esteem, and sadness. Here, higher scores indicated greater behavioral problems. Second, academic outcomes were tested, including spring reading and math IRT achievement scores and an indicator for whether or not a child would be retained in kindergarten.
For the sake of clarity, Table 7 presents fixed effects models for these five additional outcomes. There is a statistically significant main prediction of having a disability on four of the five outcomes, indicating that students with disabilities tend to have higher behavioral issues and lower achievement outcomes. That said, in no case is there a unique statistical relationship between being a student with a disability and the number of ELL classmates. This supports the mechanism presented in the introduction of this article, which provided a justification for why having a greater number of ELL classmates would be related to higher classroom social skills for students with disabilities. On the other hand, the results in Table 8 do suggest this mechanism is domain specific; that is, the number of ELL classmates does not influence problem behaviors and is not related to academic outcomes.
A final test examined the total effect of the distribution of students with disabilities and ELL classmates. Thus far, this study has examined the influence of ELL classmates on the classroom social skills outcomes of students with disabilities. Because students with disabilities and ELL classmates are in the same environment, a more complete assessment would also examine the relationship between the presence of classmates with disabilities and the same set of outcomes for ELL students in the classroom. Knowing the associations in both directions would allow administrators to make more informed decisions about the distribution of students in mainstreamed classrooms.
Table 8 presents school fixed effects models examining the relationship between the number of classmates with disabilities and the three social skills outcomes for ELL students. A student was determined as being an ELL if survey responses indicated either of the following conditions: (1) has primary household language other than English (derived from the parent survey) or (2) receives ESL instruction in a language other than English at school (derived from the teacher survey).
The model is analogous to the school fixed effects model in equation (2) and is presented as follows:
SSijk = β0 + β1Eijk*CDijk + β2Sijk + β3Fjk + β4Cjk + δk + εijkt (3)
where the key coefficient is still β1, as in previous models. Now, β1 represents the relationship between the set of outcomes and the number of classmates with disabilities (CDijk) for ELL students (Eijk) (the former being a continuous measure, and the latter being a binary measure). β1 is the interaction between an indicator for having a disability and the number of ELL classmates. The analytic rationale for utilizing an interaction term remains consistent with all previous models.
In Table 8 the first line of results presents the unique interplay between ELL students and students with disabilities in the same classroom. The coefficients are not statistically significant here. Note that the approaches of the preceding two tests of validity were tested as well, and again, no statistical significance arose. This suggests there is no relationship, neither positive nor negative, between the presence of students with disabilities and the social skills of ELL students in the same classroom.
This finding is important. While it might have been expected to see a mutually positive relationship between students with disabilities and ELL students, what is critical here is that there is no finding of negative relationships. This suggests that while a greater number of ELL students may be related to higher classroom social skills outcomes for students with disabilities, the presence of students with disabilities has no negative effects on the social skills of ELL students in the same classroom. Thus, the number of ELL classmates influenced the outcomes of students with disabilities, though there was no relationship between the number of students with disabilities and the outcomes of ELL students. This lack of a reciprocal relationship in these findings provides great insight for school administrators, who may allow one group of students to benefit by being placed with another group of students, without it detracting from the outcomes of the second group.
In examining relationship between the number of ELL classmates and the classroom social skills outcomes for students with disabilities, the present research filled multiple gaps in the literature. First, the study examined a relatively understudied aspect of peer contexti.e., having ELL students in general education classrooms. Prior to this study, little research had quantified the extent to which, if at all, ELL classmates have a spillover effect onto the outcomes of other students. Second, the present research focused on the effects of one group of high-needs students (ELL) on the outcomes of another group of high-needs students (students with disabilities) sharing a common educational space. This proved to be critical because prior research has only focused on how high-needs classmates might relate to differences in the outcomes of typical students. Thus, the study is novel in that it has examined how classroom context via peer composition relates to promoting supportive learning environments for students with disabilities. Indeed, finding ways to create positive environments for this group is certainly at the forefront of research and policy in special education.
To make these contributions to the literature, this study utilized a comprehensive dataset of kindergarten students located throughout the United States, based upon the survey responses of parents, teachers, and school administers. Given the national-level changes in the presence of ELL students and students with disabilities in classrooms across the United States, a national-level dataset was deemed appropriate. Using these data, it was possible to link student outcomes and attributes directly to the characteristics of their parents and families as well as teachers and classrooms. Thus, relying on such large-scale data allowed for the construction of empirical models to capture a greater notion of the context in which children live and learn.
Moreover, using the population of kindergarten students had two advantages, as previously mentioned. First, methodologically, children in kindergarten are placed in a single room throughout the school day, thereby ensuring that the empirical measures of the classroom do indeed capture the characteristics of the educational context in which children learn throughout the day and year. This enabled the clear identification of classroom peer groupings. Moreover, there was evidence of non-intentional sorting of students to classrooms in kindergarten in this dataset, thereby eschewing estimation issues of non-random assignment that often arise in older grades. Second, kindergarten is an exemplar for an extremely formative period, both academically and socially (Pianta & Walsh, 1996). Hence, identifying early contextual factors that promote positive social development in school has implications for both policy and practice.
Relying on these data, the methodology selected in this study was delineated by two main approaches. The first was a baseline assessment, in which each classroom social skill outcome was modeled on a wide array of observable student, family, teacher, and classroom characteristics. The second approach built directly on the first by including school fixed effects as a way to account for unobserved school-level issues that may be biasing both the key independent variable as well as dependent variables.
While the school fixed effects models were preferred from a statistical standpoint, the results were similar between both main approaches. Students with disabilities who had a greater number of ELL classmates tended to have higher frequencies of approaches to learning, interpersonal skills, and self control, relative to students with disabilities with fewer ELL classmates. These findings were supported with three tests. The first test demonstrated that even after limiting the sample to students with more experienced teachers who might have a better gauge of how to rate student social development, the coefficients were consistent with all findings. Second, in examining alternative outcomes, the results did prove to be domain-specific to classroom social skills: For students with disabilities, the number of ELL classmates did not influence problem behaviors or academic outcomes. Finally, a reverse peer effect was nonexistent. That is, the number of classmates with disabilities did not relate to the outcomes for ELL students. This final test provided insight as to how one high-needs group may have the potential to positively influence the outcomes of other high-needs classmates without detracting from their own schooling success.
These findings presented several implications. First, the data and methods of this research indicated that the presence of one high-needs group may relate positively to the outcomes of students with disabilities; here, we show this to be true as early as in kindergarten. Hence, this study provides foundational support for a positive peer effects mechanism, as delineated in the introduction of this article. Prior research in classroom peer effects has typically found positive or negative results based on how one subgroup of students (e.g., gender, ability, high needs) can influence the outcomes for the typical student in the class, thereby producing an average classroom effect. Importantly, this investigation took a new approach by examining how the presence of one high-needs subgroup of students can relate to the outcomes of another high-needs group of students in the same classroom. Hence, the research offers more in-depth insight into how the classroom context and the effects of classmates may have a unique relationship for specific high-needs groups such as students with disabilitiesa strand of research in this area that is often overlooked. School practices can thus be guided by determining not simply if one group of students performs better or worse on average, but rather by asking, better or worse for whom in particular?
Second, the research findings suggest that ELL classmates only related to classroom social skills outcomes. There was no relationship with behavioral issues, achievement, or retention. This suggests that the association of ELL classmates on the outcomes of students with disabilities is domain-specific, thereby further supporting the mechanisms laid out early in this article. Hence, the findings of the study can support educational practitioners in their efforts to create classroom settings that support specific outcomes, depending on the policy objective. While some classroom environments may promote educational attainment, the findings of this study also suggest that classroom environments may be constructed so as to support non-achievement outcomes, such as social skills. Thus, knowing which contextual factors relate to which domain of outcomes will continue to aid practitioners in making the educational adjustments necessary to maximize the influence of school settings for students with disabilities.
Third, regarding the direction of this peer contextual relationship, it was found that while the presence of ELL classmates related positively to the social skills outcomes for students with disabilities, there was no reverse effect. That is, the differences in the numbers of classmates with disabilities did not predict, neither positively nor negatively, differences in the social skills outcomes of ELL students. The findings thus imply a non-linearity, in which it is possible for group A to positively influence the outcomes of group B without having detracted from the outcomes of group A. This result has direct implications for educational practice in terms of classroom distribution; it suggests that being in a classroom with ELL students can carry advantages for students with disabilities that may not reduce the attainment of ELL students per se. It follows that resources funneled toward classrooms with both ELL classmates and students with disabilities accrue benefits across those classrooms, lending additional weight to the importance of considering the ramifications of classroom context for all students and the influences of special groups.
Finally, examining the underlying association between classroom factors and student outcomes proved to be significant even as early as kindergarten. For students with disabilities, this study shows that they have a unique relationship with others in the same classroom during this formative, first year of schooling. Consequently, the findings of the study give charge for educational practitioners and policymakers to continue identifying areas that support early classroom distributional practices and channel additional resources in a way that fosters the growth and development of all students in the general education classroom.
This research study has contributed new insight into the interplay between classroom context and classroom social skills for students with disabilities in kindergarten. Specifically, the number of ELL classmates was tested on three key classroom social skills outcomes with a large-scale dataset. The premise of this agenda is critical, as the presence of students with disabilities and ELL students continues to increase across classrooms in the United States alongside policies that are more often than not placing both demographic groups into general education classrooms. Prior to this study, empirical research had not considered how the presence of one high-needs demographic group related to differences in the socio-emotional outcomes of another high-needs group in the same educational space. For students with disabilities, a uniquely positive relationship emerged between having a greater number of ELL classmates and classroom social skills. This demonstrated the necessity for policymakers and practitioners to further address how the notion of classroom distribution along multiple capacities can promote educational development.
There are several limitations to this investigation, though each could serve as a foundation for further research. First, as described, this study is the first to examine the relationship between the presence of ELL classmates and the social skill outcomes for students with disabilities, hence contributing a new perspective to a line of research focusing on classroom context. Also as mentioned, the outcomes scales are survey-based and as such, there may be some degree of subjectivity in the ratings (see, especially, DiPerna, Lei, & Reid, 2007; Galindo & Fuller, 2010). Thus, in order to continue exploring how the classroom context may improve or detract from socio-emotional outcomes of students with disabilities, future research may explore additional child outcomes beyond achievement and development.
Second, there are many advantages to utilizing a large-scale dataset, like the ECLS-K. However, relying on other datasets might provide additional insight. For instance, in the ECLS-K dataset, it is not possible to identify traits or characteristics of ELL classmatesthe only information obtained is the aggregate classroom count available from the teacher survey. In addition, there are not large enough sample sizes to detect results by specific type of disability. In a dataset from a single school district where there is information on every student by teacher and classroom, it might be possible to delve in more detail on the characteristics of high-needs students, such as gender, race, economic status, ability, and disability. Moreover, given that the ECLS-K dataset depicts trends and patterns for the nation as a whole, it would be an important extension to examine specific schools or districts where there might be a majority of ELL students or students with disabilities. When these high-needs groups go from being minority demographic groups in national data to being the majority in these smaller school- and district-level samples, the results from the present study could be compared to those where the demographics are much different from the average U.S. demographics.
Finally, the methods presented in this study are quantitative. Consequently, the findings can be relied upon to develop conclusions based on trends and patterns. A follow-on study may also employ a qualitative approach as a way to derive more detail on the mechanism supporting the findings of this study. Assessing qualitative data from teachers and administrators in conjunction with the quantitative findings from this current study will lead to an even greater understanding of the influence of classroom context for a range of high-needs students and a range of critical outcomes.
Research for this publication was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award number 1R03HD071334. The content is solely the responsibility of the authors and does not represent the official views of the National Institutes of Health.
Aber, J. L., Bennett, N. G., Conley, D. C., & Li, J. (1997). The effects of poverty on child health and development. Annual Review of Public Health, 18, 463483.
Aizer, A. (2008). Peer effects and human capital accumulation: The externalities of ADD. NBER Working Paper Series #14354. Cambridge, MA: National Bureau of Economic Research.
Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1993). First-grade classroom behavior: Its short- and long-term consequences for school performance. Child Development, 64, 801814.
Ammermueller, A., & Pischke, J. (2006). Peer effects in European primary schools: Evidence from PIRLS. NBER working paper 12180.
Baker, E. T., Wang, M. C., & Walberg, H. J. (1994). The effects of inclusion on learning. Educational Leadership, 52, 3335.
Banks, J. A., & Banks, C. A. M. (1995). Handbook of research in multicultural education. New York, NY: Macmillan.
Blackorby, J., & Wagner, M. (1996). Longitudinal postschool outcomes of youth with disabilities: Findings from the National Longitudinal Transition Study. Exceptional Children, 62, 399414.
Boyle, C. A., Boulet, S., Schieve, L. A., Cohen, R. A., Blumberg, S. J., Yeagrin-Allsopp, M., . . . Kogan, M. D. (2011). Trends in the prevalence of developmental disabilities in U.S. children, 1997-2008. Pediatrics, 127, 10341042.
Brooks-Gunn, J., & Duncan, G. (1997). The effects of poverty on children. The Future of Children, 7, 5571.
Burke, M. A., & Sass, T. R. (2008). Classroom peer effects and student achievement. Calder, Working Paper 18.
Carlberg, C., & Kavale, K. (1980). The efficacy of special versus regular class placement for exceptional children: A meta-analysis. The Journal of Special Education, 14, 459462.
Cho, R. M. (2012). Are there peer effects associated with having English Language Learner (ELL) classmates? Evidence from the Early Childhood Longitudinal Study Kindergarten Cohort (ECLS-K). Economics of Education Review, 31, 629643.
Coplan, R. J., & Prakash, K. (2003). Spending time with teacher: Characteristics of preschoolers who frequently elicit versus initiate interactions with teachers. Early Childhood Research Quarterly, 18, 143158.
De Cohen, C., Deterding, N., & Clewell, B. (2005). Whos left behind? Immigrant children in high- and low-LEP schools. Washington, DC: The Urban Institute.
Dee, T. S., & West, M. R. (2012). The non-cognitive returns to class size. Educational Evaluation and Policy Analysis, 33, 2346.
Demaray, M. K., Ruffalo, S. L., Carlson, J., Busse, R. T., & Olson, A. E. (1995). Social skills assessment: A comparative evaluation of six published rating scales. School Psychology Review, 24, 648671.
DiPerna, J. C., Lei, P. W., & Reid, E. E. (2007). Kindergarten predictors of mathematical growth in the primary grades: An investigation using the Early Childhood Longitudinal Study Kindergarten Cohort. Journal of Educational Psychology, 99, 369379.
Drotar, D. (1997). Relating parent and family functioning to the psychological adjustment of children with chronic health conditions: What we have learned? What do we need to know? Journal of Pediatric Psychology, 22, 149165.
Fletcher, J. (2010). Spillover effects of inclusion of classmates with emotional problems on test scores in early elementary school. Journal of Policy Analysis and Management, 29(1), 6983.
Fry, R. (2008). The role of schools in the English Language Learner achievement gap. Washington, DC: Pew Hispanic Center.
Fryer, R., & Levitt, S. (2004). Understanding the black-white test score gap in the first two years of school. The Review of Economics and Statistics, 86, 447464.
Galindo, C., & Fuller, B. (2010). The social competence of Latino kindergartners and growth in mathematical understanding. Developmental Psychology, 46, 579592.
Gottfried, M.A. (2014a). Classmates with disabilities and students non-cognitive outcomes. Educational Evaluation and Policy Analysis, 36, 20-43.
Gottfried, M.A. (2014b). The positive peer effects of diversity: Exploring the relationship between English Language Learner classmates and socio-emotional skills in early elementary school. Elementary School Journal, 115, 22-48.
Greshan, F. M., & Elliott, S. N. (1990). Social skills rating system. Circle Pines, MN: American Guidance Service.
Gullo, D. F. (2000). The long-term educational effects of half-day vs. full-day kindergarten. Early Child Development and Care, 160, 1724.
Hanushek, E., Kain, J., Markman, J., & Rivkin, S. (2003). Does peer ability affect student achievement? Journal of Applied Econometrics, 18, 527544.
Hanushek, E., Kain, J., & Rivkin, S. (2002). Inferring program effects for special populations: Does special education raise achievement for students with disabilities? Review of Economics and Statistics, 84(4), 584599.
Hayworth, P. (2009). The quest for a mainstream EAL pedagogy. Teachers College Record, 111, 21792208.
Hazel, J. S., & Shumaker, J. B. (1988). Social skills and learning disabilities: Current issues and recommendations for future research. In J. F. Kavanah & T. J. Truss (Eds.), Learning disabilities proceedings of the national conference (pp. 293344). Parkton, MD: York Press.
Hoxby, C. (2000). The effects of class size on student achievement: New evidence from population variation. Quarterly Journal of Economics, 115, 12391285.
Ingersoll, R. (2002). The teacher shortage: A case of wrong diagnosis and wrong prescription. NASSP Bulletin, 86, 1630.
Juel, C. (1988). Learning to read and write: A longitudinal study of 54 children from first through fourth grades. Journal of Educational Psychology, 80, 437447.
Karabenick, S., & Noda, P. (2004). Professional development implications of teachers beliefs and attitudes toward English Language Learners. Bilingual Research Journal, 28, 5575.
Karweit, N. (1992). The kindergarten experience. Educational Leadership, 49, 8286.
Kavale, N. A., & Forness, S. R. (1995). The nature of learning disabilities: Critical elements of diagnosis and classification. Mahweh, NJ: Erlbaum.
Kavale, K. A., & Mostert, M. P. (2004). Social skills interventions for individuals with learning disabilities. Learning Disability Quarterly, 27, 3143.
Konold, T. R., & Pianta, R. C. (2007). The influence of informants ratings of childrens behavioral functioning: A latent variable approach. Journal of Psychoeducational Assessment, 25, 222236.
Kontos, S., & Wilcox-Herzog, A. (1997). Teachers interactions with children: Why are they so important? Young Children, 52, 412.
Lee, V. E., Burkam, D. T., Ready, D. D., Honigman, J., & Meisels, S. J. (2006). Full-day versus half-day kindergarten: In which program do children learn more? American Journal of Education, 112(2), 163208.
Lee, R., Zhai, F., Brooks-Gunn, J., Han, W. J., & Waldfogel, J. (2013). Head Start participation and school readiness: Evidence from the Early Childhood Longitudinal StudyBirth Cohort. Developmental Psychology 50(1), 20215.
Lindsay, G. (2007). Educational psychology and the effectiveness of inclusive education/mainstreaming. British Journal of Educational Psychology, 77(1), 124.
Lipsky, D. K., & Gartner, A. (1995). National study on inclusion: Overview and summary report. In National Center on Educational Restructuring and Inclusion Bulletin, 2(2), 18.
Loeb, S., Bridges, M., Bassok, D., Fuller, B., Rumberger, R. W. (2007). How much is too much? The influence of preschool centers on childrens social and cognitive development. Economics of Education Review, 26, 5266.
Luekens, M. T., Lyter, D. M., & Fox, E. E. (2004). Teacher attrition and mobility: Results from the Teacher Follow-up Survey, 2000-01 (NCES 2004-301). US Department of Education, National Center for Education Statistics. Washington, DC: Government Printing Office.
Meisels, S. J., Atkins-Burnett, S., & Nicholson, J. (1996). Assessment of social competence, adaptive behaviors, and approaches to learning with young children. Washington, DC: U.S. Department of Education, Office of Educational Research and Improvement.
Moon, N., Todd, R., Morton, D., & Ivey, E. (2012). Accommodating Students with Disabilities in Science, Technology, Engineering, and Mathematics (STEM): Findings from Research and Practice for Middle Grades through University Education. Atlanta, GA: Center for Assistive Technology and Environmental Access, College of Architecture, Georgia Institute of Technology.
NCELA. (2010). The growing numbers of English learner students 1997/98-2007/08. Accessed at: http://www.ncela.gwu.edu/files/uploads/9/growingLEP_0708.pdf
National Center for Education Statistics (U.S. Department of Education). (2011). Students with disabilities. Digest of Education Statistics, 2010 (NCES 2011-015), Chapter 2. Washington, DC: National Center for Education Statistics.
National Education Association. (2008). English language learners face unique challenges. Washington, DC: National Education Association.
Neidell, M., & Waldfogel, J. (2010). Cognitive and noncognitive peer effects in early education. The Review of Economics and Statistics, 92, 562576.
Olson, S. L., Sameroff, A. J., Kerr, D. C. R., Lopez, N. L., & Wellman, H. M. (2005). Developmental foundations of externalizing problems in young children: The role of effortful control. Developmental and Psychopathology, 17, 2545.
Phelps, L. A., & Hanley-Maxwell, C. (1997). School to work transitions for youth with disabilities: A review of outcomes and practices. Review of Educational Research, 67, 197226.
Pianta, R., & Walsh, D. (1996). High-risk children in schools: Constructing sustaining relationships. New York, NY: Routledge.
Posner, M. I., & Rothbart, M. K. (2000). Developing mechanisms of self-regulation. Development and Pscyhopathology, 12, 427441.
Reeves, J. (2006). Secondary teacher attitudes toward including English Language Learners in mainstream classrooms? The Journal of Education Research, 99, 131142.
Rhum, C. J. (2004). Parental employment and child cognitive development. Journal of Human Resources, 39, 155192.
Rule, A. C., Stefanich, G., Haselhuhn, C., & Peiffer, B. (2009). A Working Conference on Students with Disabilities in STEM Coursework and Careers (Abstract). Iowa City, IA: University of Iowa, College of Education.
Schneider, B., Carnoy, M., Kilpatrick, J., Schmidt, W. H., & Shavelson, R. J. (2007). Estimating causal effects using experimental and observational designs. Washington, DC: American Education Research Association.
Smith, S. (1997). A longitudinal study: The literacy development of 57 children. In C. Kinzer, K. Hinchman, & D. Leu (Eds.), Inquiries in literacy theory and practice (pp. 250264). Chicago, IL: National Reading Conference.
Supalo, C. A., Mallouk, T. E., Rankel, L., Amorosi, C., & Graybill, C. M. (2008). Low-cost laboratory adaptations for precollege students who are blind or visually impaired. Journal of Chemical Education, 85, 243247.
Thurlow, M. L., Quenemoen, R. F., & Lazarus, S. S. (2012). Leadeship for student performance in an era of accountability. In J. B. Crockett, B. S. Billingsley, & M. L. Boscardin (Eds.), Handbook of leadership and administration for special education (pp. 316). New York, NY: Routledge.
Tichenor, M. S., Heins, B., & Piechura-Couture, K. (2000). Parent perceptions of a co-taught inclusive classroom. Education, 120, 569574.
U.S. Department of Education. (2007). Demographic and school characteristics of students receiving special education in the elementary grade. Washington, DC: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education, NCES 2007-0005.
U.S. Department of Education, National Center for Education Statistics. (2012). Digest of Education Statistics, 2011. Washington, DC: U.S. Department of Education.
Walston, J., & West, J. (2004). Full-day and half-day kindergarten in the United States: Findings from the Early Childhood Longitudinal Study, Kindergarten Class of 1998-1999. Washington, DC: National Center for Education Statistics.
Waterman, C., McDermott, P. A., Fantuzzo, J. W., & Gadsden, V. L. (2012). The matter of assessor variance in early childhood education Or whose score is it anyway? Early Childhood Research Quarterly, 27, 4654.
Weast, J. D. (2004). Early Success: Closing the Opportunity Gap for our Youngest Learners. Rockville, MD: Montgomery County Public Schools.
Wiley, T., & Wright, W. (2004). Against the undertow: Language-minority education policy and politics in the age of accountability. Educational Policy, 18, 142168.
Williams, L. J., & Downing, J. E. (1998). Membership and belonging in inclusive classrooms: What do middle school students have to say? Journal of the Association for Persons with Severe Handicaps, 23(2), 98110.
Winters, M. A., & Greene, J. P. (2007). Debunking a special education myth. Education Next, 7, 6771.
Yamauchi, C., & Leigh, A. (2011). Which children benefit from non-parental care? Economics of Education Review, 30, 14681490.
Zimmer, R. W., & Toma, E. F. (2000). Peer effects in private and public schools across countries. Journal of Policy Analysis and Management, 19, 7592.