April 2, 2008
Achieving Equity in Special Education: History, Status, and Current Challenges
By Skiba, Russell J Simmons, Ada B; Ritter, Shana; Gibb, Ashley C; Rausch, M Karega; Cuadrado, Jason; Chung, Choong-Geun
ABSTRACT: Among the most-longstanding and intransigent issues in the field, the disproportionate representation of minority students in special education programs has its roots in a long history of educational segregation and discrimination. Although national estimates of disproportionality have been consistent over time, state and heal estimates may show varying patterns of disproportionality. A number of factors may contribute to disproportionality, including test bias, poverty, special education processes, inequity in general education, issues of behavior management, and cultural mismatch/cultural reproduction. This article provides a report on the history, measurement, status, and factors contributing to disproportionate representation in special education, and offers recommendations based on an understanding of racial and ethnic disparities in special education as a multiply determined phenomenon. Special education was borne out of, and owes a debt to, the civil rights movement. That is, the inspiration for, and the strategies used by, advocates whose efforts resulted in the first national special education legislation emerged from the struggles of the civil rights movement (Smith & Kozleski, 2005). Concerns about racial inequity were central to litigation (e.g., Mills v. Board of Education, 1972) that led to the promulgation of the first special education legislation (Individuals With Disabilities Education Act, IDEA, Public Law No. 94142, 1975). Thus, it is highly ironic that racial disparities in rates of special education service remain one of the key indicators of inequity in our nation's educational system.The disproportionate representation of minority students is among the most critical and enduring problems in the field of special education. Despite court challenges (Larry P. v. Riles, 1972/1974/ 1979/1984; PASEv. Hannon, 1980); federal reports (Donovan & Cross, 2002; Heller, Holtzman, & Messick,1982); and abundant research on the issue (e.g., Chinn & Hughes, 1987; Harry & Klingner, 2006; Hosp & Reschly, 2003; Losen & Orfield, 2002; Oswald, Coutinho, Best, & Singh, 1999), the problem of disproportionate representation of minority students in special education has persisted. Indeed, although consistently documented, it is fair to say that the full complexity of minority disproportionality has not yet been understood, nor has a clear or comprehensive picture emerged concerning the causes of disproportionality (Donovan & Cross; Harry & Klingner). To address the issue of disproportionate minority placement, the 1997 reauthorization of the Individuals With Disabilities Education Act (IDEA 97, Public Law No. 105-17) stressed the importance of efforts to "prevent the intensification of problems connected with mislabeling and high dropout rates among minority children with disabilities" (p. 5) and that effort has been further amplified in the Individuals With Disabilities Education Improvement Act (IDEA 2004, Public Law No. 108-446).
This article provides a status report on minority disproportionality in special education. What is the historical context for current problems of racial/ethnic disparity? What are the current levels of disproportionality and how are those measured? What are the possible causes and conditions that create or maintain disproportionality? What interventions have been suggested? Finally, the history and current status of the field suggests that any comprehensive strategy for addressing disproportionality must attend to three aspects of the issue: (a) examination of current data, (b) comprehensive hypothesis formulation and interpretation, and (c) culturally responsive intervention and evaluation.
HISTORY: A BRIEF SYNOPSIS OF A VERY OLD PROBLEM
The initial identification of the problem of disproportionate representation of some groups, most notably African American students, in special education is often traced back to Dunn's (1968) classic critique of the field. Yet the problem itself has its roots far deeper, in the problems of oppression and discrimination that have characterized race relations throughout American history (Smedley, 2007). In 1853, Margaret Douglas was sentenced to 1 month in jail for her attempts to teach the children of freed slaves to read and write (Blaustein & Zangrando, 1968). In 1896, Plessy v. Ferguson legitimated the doctrine of separate but equal, even though segregated education in the Jim Crow period was by no means equal (Jackson & Weidman, 2006). In the late 19th century and early 20th century, attacks on Black communities during race riots included the burning of Black schools (Harmer, 2001). Early 20th century mental testing was grounded in the premise of American eugenics that races other than those of northern European stock were intellectually inferior, and that the purity of the superior races should be preserved by vigorously segregating the "feeble-minded" (Terman, 1916). From Reconstruction until the 1950s, the dominant view of African American education was that it was intended not to educate for equal citizenship, but rather for the lower ranked positions that it was assumed African Americans would occupy (Rury, 2002).
It is not surprising then that leaders in the emerging field of special education documented racially-based disparities in service in the 1960s and 1970s. In his classic critique of special education, Dunn (1968) suggested that the overrepresentation of ethnic and language minority students in self-contained special education classrooms raised significant civil rights and educational concerns. Mercer (1973), highlighting ethnic differences in rates of special education service as part of her critique of the "6-hour" or situationally retarded child, found that public schools tended to identify more children as mentally retarded rhan any other child service setting.
In the wake of Brown v. Board of Education (1954) and legislative action to provide equal access to education, institutional structures, such as ability grouping and significantly separate special education classrooms, continued to keep minority students segregated from their White peers (Losen & Welner, 2001). Addressing violations of the Equal Protection Clause of the Constitution and Title VI of the Civil Rights Act of 1964, de facto segregarion was challenged in the Washington, DC public school system in the case of Hobson v. Hansen (1967). Continued challenges were brought in court under Title VI of the Civil Rights Act of 1964, the Rehabilitation Act of 1973, and the Education for All Handicapped Children Act of 1975, addressing the role of standardized testing and the reduced educational opportunity afforded by the racial isolation of minorities in special education programs (Diana v. California State Board of Education, 1970; Guadalupe Organization v. Temple Elementary School District #3, 1972; Larry P. v. Riles, 1972/1974/ 1979/1984; RASE v. Hannon, 1980). Although the earliest of these cases were highly influential in the generation of state and federal statutes establishing special education in the early to mid-1970s, aie outcomes of the cases were by no means uniform (Bersoff, 1981; Reschly, 1996). Nevertheless, concerns about bias in testing led to a profusion of research in the 1970s and early 1980s examining that issue.
In the 1980s, examination of the U.S. Department of Education Office for Civil Rights survey data began to produce estimates of the extent and distribution of disproportionality, which have been consistent over time (Chinn & Hughes, 1987; Donovan & Cross, 2002; Finn, 1982). Yet this research did not, in and of itself, provide any understanding of the mechanisms that contribute to racial and ethnic disparities in special education. Recent disproportionality research has seen a sharper focus on the forces that shape and maintain disproportionate representation (e.g., Artiles, 2003; Harry & Klingner, 2006; Hosp & Reschly, 2003; Skiba et al., 2006a).
Policy pressure to remediate disproportionality in special education at the state and local levels increased significantly with the inclusion of provisions concerning disproportionality in IDEA 1997 and especially with the expansion of provisions in the reauthorization of IDEA in 2004 (see Figure 1). Under the provisions of IDEA 2004, states must monitor disproportionate representation by race or ethnicity in disability categories and special education placements and require the review of local policies, practices, and procedures when disproportionate representation is found. One of rhe most significant new requirements under IDEA 2004 is that local educational agencies (LEAs) determined to have significant disproportionality must devote the maximum amount of Part B funds allowable (15%) to early intervening programs. Early intervening services are distinguished from early intervention services for infants and toddlers with disabilities in that they identify and target "children who are struggling to learn . . . and quickly intervening to provide support" (Williams, 2007, p. 28). Significant disproportionality is not defined in IDEA 2004 nor its implementing regulations, and discretion is left to the states to develop the quantifiable indicators of disproportionality used for determining significance. MEASUREMENT ISSUES IN DISPROPORTIONALITY
Disproportionality may be defined as the representation of a group in a category that exceeds our expectations for that group, or differs substantially from the representation of others in that category. Although concerns have historically tended to focus on issues of overrepresentation in special education or specific disability categories, groups may also be underrepresented in a category or setting (e.g., underrepresentation in general education settings, gifted education, or visual impairment). Although the concept of disproportionate representation seems relatively straightforward, measurement of disproportionality can be quite complex. In measuring disproportionality, one may assess (a) the extent to which a group is over- or underrepresented in a category compared to its proportion in the broader population (composition index) or (b) the extent to which a group is found eligible for service at a rate differing from rhat of other groups (risk index and risk ratio).
The most intuitive method of measurement of disproportionality, the composition index (CI; Donovan & Cross, 2002), compares the proportion of those served in special education represented by a given ethnic group with the proportion that group represents in the population or in school enrollment; that is, it provides a measure of representation in the target phenomenon compared to our expectations for that group. At the national level, African American students account for 33% of students identified as mentally retarded, clearly discrepant from their representation in the school- age population of 17% (Donovan & Cross).
Although the CI is a clear cut measure, there are some difficulties with its use. First, there is no criterion for determining when a discrepancy in composition indices is meaningful or significant (Coutinho & Oswald, 2004). Chinn and Hughes (1987) suggested setting a confidence level of 10% around the population enrollment percentage of the group in question (e.g., for an overall African American enrollment of 17%, disproportionality would be expressed by special education enrollment rates outside of a range of 17% +/1.7%, that is, 15.3% to 18.7%). The CI is also beset by scaling problems: discrepancies at the extremes of the scale may not have the same meaning as those in the middle. Finally, the CI diminishes in usefulness as groups become more homogeneous (Westat, 2003, 2005). In several urban settings, African American enrollment exceeds 92%, making it impossible to find overrepresentation (e.g., 92% + 9.2% = 101.2% using Chinn & Hughes' criteria).
RISK INDEX AND RELATIVE RISK RATIO
An alternative approach to describing disproportionality is to measure a group's representation in special education compared to other groups. The risk index (RI) is the proportion of a given group served in a given category and represents the best estimate of the risk for that outcome for that group. Donovan and Cross (2002) reported, for example, that, at the national level, 2.64% of all African American students enrolled in the public schools are identified as having mental retardation (MR). By itself, however, the RI is not particularly meaningful. In order to interpret the RI, a ratio of the risk of the target group to one or more groups may be constructed, termed a risk ratio (RR; Hosp & Reschly, 2003; Parrish, 2002). A ratio of 1.0 indicates exact proportionality, whereas ratios above or below 1.0 indicate overand underrepresentation, respectively. Comparing African American risk for MR identification (2.64%) with the risk index of 1.18% of White students for that disability category yields a risk ratio of 2.24 (2.64/1.18), suggesting that African Americans are more than two times more likely to be served in the category mental retardation than White students. The same data can also be used to compute an odds ratio, representing both the probability of being in special education and the probability of not being in special education for both groups (Finn, 1982). In contrast to the RR, the odds ratios assess both occurrence and nonoccurrence data.
Provisions of IDEA 2004 With Respect to Minority Disproportionality in Special Education
There are also limitations and issues of interpretation with the RR. Although less sensitive to changes in relative proportions of population, the RR may become unstable with small >>'s (Hosp & Reschly, 2004). Risk ratios may also provide an incomplete picture of racial and ethnic disparities; although both 30% of Blacks versus 15% of Whites in a category will provide the same RR (2.0) as 2% of Blacks and 1% of Whites in that category, the meaning of those discrepancies varies greatly. Finally, there is no consensus in the field on the appropriate group against which to compare a target group's RI. A case can be made that, being the largest and historically dominant group, White enrollment represents the appropriate criterion against which to compare other racial/ethnic group representation and may be a more appropriate measure for assessing Latino disproportionality. Using White as the index group precludes the calculation of a RR for that group, however, making estimation of White underrepresentation in special education impossible (Westat, 2004). The U.S. Department of Education, Office of Special Education Programs recommends using all others as the denominator in the calculation of disproportionality (Westat, 2005), but the use of either Whites and All Others as the index group appears ro be acceptable in the research literature (Skiba, Poloni- Staudinger, Simmons, Feggins, & Chung, 2005).
In order to aid states in the reporting of disproportionality data, the U.S. Department of Education, Office of Special Education Programs and Westat convened a national panel to consider methodologies for monitoring disproportionality. The guidance developed as a result of that panel (a) recommends the use of a RR approach to measure disproportionality; (b) provides instruction on the calculation of those measures; and (c) recommends an alternative "weighted" RR when there are fewer than 10 students from a target group in a given school district, or to compare RRs across districts (Westat, 2004, 2005). Again, absolute criteria for significant disproportionality are left undefined.
Although there has been progress in recent years in standardizing the measurement of disproportionality, significant areas of confusion remain. Although different measures such as RRs and odds ratios are sometimes equated or confused in the literature (see e.g., Donovan & Cross, 2002), they provide similar data only under certain conditions (Davies, Crombie, & Tavakoli, 1998). Further, the issue of a definitive criteria in determining disproportionality is complex. The ffamers of IDEA 2004 may have deliberately intended to avoid cutoffs identifying significant disproportionality in order to allow responsiveness to regional and local variation; rigidly defined criteria might also encourage local districts to meet those criteria by simply cutting minority referrals. Yet, the absence of criteria for defining significant disproportionality may perpetuate confusion by failing to provide sufficient guidance to those at the state and local level who may be unfamiliar with statistical analysis.
STATUS OF DISPROPORTIONALITY
PATTERNS OF DISPROPORTIONALITY
Analyses of data from the U.S. Department of Education, Office for Civil Rights (OCR; e.g., Chinn & Hughes, 1987; Donovan & Cross, 2002; Finn, 1982) have revealed consistent patterns of disproportionality. African American students are typically found to be overrepresented in overall special education service and in the categories of mental retardation (MR) and emotional disturbance (ED), whereas American Indian/Alaska Native students have been overrepresented in the category of learning disabilities (LD). Data from the 26th Annual Report to Congress on the Implementation of the Individuals With Disabilities Education Act (U.S. Department of Education, 2006; see Table 1) indicates that American Indian/Alaska Native students received services under the category developmental delay at a higher rate than other groups, Asian/Pacific Islander students received special education for hearing impairments and autism at a somewhat higher rate than other students, and Latino students were somewhat more likely to receive services in the category of hearing impairment. Parrish (2002) reported that African American students are the most overrepresented group in special education programs in nearly every state.
Risk Ratios for All Disability Categories and Racial/Ethnic Categories From the 26th Annual Report to Congress
A number of characteristics of disproportionality have been noted. Disproportionate representation is greater in the judgmental or "soft" disability categories of MR, ED, or LD than in the nonjudgmental or "hard" disability categories, such as hearing impairment, visual impairment, or orthopedic impairment (Donovan & Cross, 2002; Parrish, 2002). Parrish reported that rates of overrepresentation tend to increase as a minority group constitutes a relatively high percentage of their states' population. Finn (1982) reported a complex relationship between school district size and percentage of minority enrollment-for smaller districts, disproportionality was greatest in districts with the highest minority enrollment, whereas for larger districts (30,000 or more students), disproportionality was greatest when minority enrollment was 30% or less. Finally, states may show evidence of disproportionality in categories that appear proportionate at the national level, and local school districts may show evidence of disproportionality in a category not disproportionate at the state level (Harry & Klingner, 2006). In contrast to the relative stability of African American disproportionality over time, there have been inconsistencies in estimates of the degree and direction of Latino disproportionality. Some state- and district-based studies, primarily based on data from California or New York, have tended to show Latino overrepresentation in special education (Artiles, Rueda, Salazar & Higareda, 2002; Wright & Santa Cruz, 1983). National data, however, show that the most common finding is the underrepresentation of Latino students in overall special education service and in most disability categories (Chinn & Hughes, 1987; National Center on Culturally Responsive Educational Systems, NCCRESt, 2006). Examination of Table 1 suggests Hispanic overrepresentation in Hearing Impairments and perhaps LD; underrepresentation is a much more common finding across a number of disability categories.
Discrepancies between findings of overrepresentation for African American students and underrepresentation for Latino students may be due in part to the tendency for overrepresentation to become more pronounced as minority students represent a larger proportion of the population. In contrast to the case of African American students, where overrepresentation in certain categories has been found to be relatively consistent across time and locale, overrepresentation of Latino students appears to be concentrated in those areas in which Latino students represent a relatively higher proportion of enrollment (Parrish, 2002). Formal studies to evaluate these discrepancies have been limited (Klingner, Artiles, & Mendez Barletta, 2006). The difficulty in accurately distinguishing between language acquisition difficulties for English Language Learners and a language disability also complicates identification and assessment issues for Latino students (Barrera, 2006; Ortiz, 1997).
DISPROPORTIONALITY IN EDUCATIONAL SETTINGS
Although less well researched, available data demonstrates that students of color, especially African Americans, are overrepresented in more restrictive educational environments and underrepresented in less restrictive settings (Fierros & Conroy, 2002; Skiba, Poloni- Staudinger, Gallini, Simmons, & Feggins-Azziz, 2006b). Given the conceptual importance of inclusion and the dramatic increases in recent years in general education placements for students with disabilities (McLeskey, Henry, & Axelrod, 1999), it could be argued that disproportionality with respect to access to less restrictive educational environments may be more important conceptually than disparities in disability category (Skiba et al., 2006b).
Different interpretations might well be applied to findings of racial disparities in educational settings. It might be presumed, for example, that "differences in placement by race/ethnicity may reflect the disproportional representation of some minority groups in disability categories that are predominately served in more restrictive settings" (U.S. Department of Education, 2002, p. Ill- 45). Yet failure to find such a pattern may suggest that disproportionality in special education settings is driven, to some extent, by systemic responses, such as educators who may mistake cultural differences for cognitive or behavioral disabilities (Harry, 2008; Oswald et al., 1999; Trent, Kea, & Oh, 2008).
To test that hypothesis, Skiba et al. (2006b) explored the extent to which African American students were proportionately placed in more and less restrictive settings within five disability categories in one state's data for a single year. In four of the five disability categories, African American children were more likely than their peers with the same disability to be overrepresented in more restrictive settings, or underrepresented in the general education setting. Further, disproportionality in placement increased as the severity of the disability decreased: African American students with disabilities were much more likely than peers with the same disability label to be served in a separate class setting in milder, more judgmental categories such as learning disabilities (RR = 3.20) or speech and language (RR = 7.66). Such results do not support the hypothesis that minority disproportionality in educational environments is simply a function of disproportionality in disability category. That is, the overuse of more restrictive placements for African American students with disabilities is likely due to factors other than severity of disability; further research is critically needed to identify what those factors might be.
CAUSES OF DISPROPORTIONATE SPECIAL EDUCATION REPRESENTATION
A fairly extensive database has consistently documented African American disproportionality in special education service and across educational environments, although findings regarding Latino disproportionality are less extensive and less consistent. Describing the extent of the problem is merely the first step in understanding the causes and conditions that create and maintain racial disparities in special education. A number of possible conditions or causes related to special education disproportionality have been explored, beginning in the 1970s with test bias.
PSYCHOMETRIC TEST BIAS
In the 1970s, the issue of psychometric test bias played a central role in court cases concerning minority disproportionality, specifically overrepresentation. These cases appeared to be based on the presumption that tests that yielded group racial differences in results must, of necessity, be biased (Mercer, 1973). Although the presiding judge in Larry P. v. Riles (1972/1974/1979/1984) appeared to agree with this assessment, other courts failed to find evidence that bias in assessment has yielded misclassification (Bersoff, 1981). The possibility of bias against minorities in standardized tests of intelligence and achievement was examined fairly extensively in the 1970s and 1980s, although there has been less research on the topic in recent years (Valencia & Suzuki, 2000), focusing mainly on the impact of highstakes testing (Madaus & Clarke, 2001). Extensive reviews of that literature have reached somewhat different conclusions.
Perhaps the most influential review of cultural bias in psychometric tests was conducted by Jensen (1980). That review and others (Brown, Reynolds, & Whitaker, 1999; Cole, 1981) concluded that data from a number of converging sources indicates little or no evidence of bias against minority students in intelligence tests. First, a similar factor srructure for intelligence tests for Black and White students suggests that the major constructs underlying those tests are comparable across ethnic groups (Brown et al.). second, although it has been argued that undersampling of minority populations will lead to tests that are biased against minority populations (Harrington, 1975), tests of the hypotheses with human samples have not yielded such results (Hickman & Reynolds, 1987). Finally, comparisons of African American and White performance on a wide range of tests have generally failed to find a significant bias at the item level (Brown et al.). For diese reasons, it has been argued that the case against test bias has been conclusively made (Jensen) and some have expressed frustration about the failure of the field to fully accept such findings (Reynolds, 2000).
Other equally extensive reviews of the same literature have not always reached the same conclusions, however. Valencia and Suzuki (2000) noted that, because the majority of studies on test bias were conducted in the 1970s and 1980s, almost all of what we know regarding test bias is based on the WISC and WISC-R intelligence tests, neither of which is currently in use. Further, the literature on test bias has underrepresented students in special education and some minority groups. Nor are the results of available research entirely consistent. Of 32 investigations of content and predictive bias reviewed by Valencia and Suzuki, 50% yielded findings concerning bias that were at least mixed; in the area of predictive validity, 6 out of 18 investigations (involving primarily Mexican Americans, but also African Americans and Asian Americans) showed evidence for bias in predictive validity.
In particular, recent research has pointed to possible sources of item bias. Shepard (1987), arguing that analysis at the individual item level may be insufficient for exploring test bias, suggested that more sophisticated methodologies, such as item response theory, have yielded patterns of bias that explain a small but significant portion of the variance in Black-White test score discrepancies. In particulat, concerns have been raised in regard to item selection processes on commercially available standardized tests that may be weighted differentially against minority test takers (Freedle, 2003; Kidder & Rosner, 2002). Examining the test construction process for the SAT, Kidder and Rosner found that questions more frequently answered correctly by African American students than White students are rejected at a higher frequency for inclusion, because such items do not correlate with a total score that is higher for White than Black test takers. Further research is necessary to determine to what extent such processes may apply in the construction of standardized tests of intelligence or achievement used in special education assessment.
Finally, language differences and examiner effects may also contribute to bias in testing. Abedi (2004) demonstrated that tests normed for native English speakers have lower reliability and validity for English Language Learners and noted that tests standardized on native English speakers may inadvertently function as English language proficiency tests. The examiner may also represent a source of bias. In a meta analysis of the effects of examiner familiarity on test performance, Fuchs and Fuchs (1986) reported that examiner unfamiliarity, defined in part as membership in a different group from the examinee, had a significant impact on standardized test performance. In particular, the examinees of low socioeconomic status (SES) were more significantly affected than examinees of higher SES. SUMMARY
An extensive literature exploring psychometric test bias has, in general, tended not to identify a level of cultural bias in standardized tests of intelligence sufficient to account for the inappropriate classification of students as disabled. Yet, given the failure to include relevant populations in some areas of study, a literature base that is, for the most part, more than 20 years old, and inconsistent evidence in certain areas (e.g., item bias, examiner bias), the assertion that test bias has been conclusively ruled out as a possible source of minority disproportionality in special education is at best premature.
Even a demonstration rhat standardized tests of cognition were completely free of psychometric bias would not in and of itself identify die source of the Black-White test score gap; in particular, findings rhat tests are unbiased does not mean that racial differences in IQ scores are inherent or genetic. Tests that are technically unbiased may still provide an index that is essentially still unfair to certain groups if interpreted uncritically. Wide and consistent disparities have been identified in the quantity and quality of educational resources available to White and African American students in American education (Donovan & Cross, 2002; Kozol, 2005). While depressed minority test scores are an indicator of current performance, they are also a product rhat accurately reflects the impact of economic and educational disadvantage. Tests that are unbiased may provide an accurate estimate of current individual aptitude; yet they also provide an unbiased and accurate record of the effects of unequal educational opportunity. Indeed, at this point in history, tests that failed to reflect some form of disadvantage for victims of racial or socioeconomic bias might be said to lack concurrent and predictive validity with respect to the conditions of bias present in our educational and social systems (Skiba, Bush & Knesting, 2002).
SOCIO-DEMOGRAPHIC FACTORS: THE INFLUENCE OF POVERTY
A second factor that might contribute to a disproportionate rate of representation in special education among students of color are sociodemographic factors associated with economic disadvantage. One might expect that because minority students are more likely to be exposed to poverty in American society (U.S. Census Bureau, 2001), the risk factors associated with poverty will result in increased academic underachievement and emotional/behavioral problems among minority students, thus increasing the risk of minority referral to special education.
A number of demographic factors related to geographical location and SES have been shown to be associated wirh student educational achievement or early cognitive development. These include neighborhood and housing stability (Ainsworth, 2002); the student's home environment (Caldas & Bankston, 1999); family health care (Kramer, Allen, & Gergen, 1995); and geographic location (Huebner, 1985). McLloyd (1998) reported that the effects of poverty on early cognitive development, school achievement, and socio-emotional functioning are dependent on the duration, timing, and neighborhood context of poverty; deep and persistent poverty consistently predicts more deleterious effects. The 2002 National Research Council panel exploring disproportionality in special education (Donovan & Cross, 2002) affirmed that biological and social/ environmental factors that disproportionately affect minority students have been found to contribute to poor cognitive and behavioral outcomes, and they recommended a national commitment to early intervention to offset socioeconomic risk factors.
The consistent overlap of race and poverty in this country has led some to suggest that race is simply a "proxy" for poverty (Hodgkinson, 1995). MacMillan and Reschly (1998) argued that the correlation of ethnicity and social class suggests that class may explain more variance than race in predicting service in high- incidence disabilities. That view is also widely shared among school personnel (see e.g., Harry, Klingner, Sturges, & Moore, 2002; Skiba et al., 2006a).
Yet showing that poverty influences academic achievement is not the same thing as demonstrating that poverty causes minority disproportionality in special education, Skiba et al. (2005) noted that developing a link between poverty and minority disproportionality requires a series of logical connections, not all of which are well-documented in the literature. Although there is a fairly strongly documented connection between minority status and poverty (U.S. Census Bureau, 2001), direct links between poverty and academic and behavioral outcomes are not as impressive (Brooks-Gunn & Duncan, 1997). Nor do academic or social/behavioral problems necessarily predict special education eligibility, because the specific disability definitions of IDEA are intended to ensure that not all students with academic or emotional/behavioral problems are found eligible for special education. Thus, to demonstrate rhat poverty contributes significantly to special education disproportionality, it would be necessary to show that economic disadvantage increases the risk, not merely of underachievement, but of the specific types of learning and behavior problems defined by IDEA as disability.
Given this complexity, it is not surprising that investigations of the association of poverty and special education disproportionality have yielded inconsistent results that sometimes contradict the race-poverty hypothesis. Some have found that poverty indeed creates higher rates of minority placement in the disability categories of LD (Coutinho, Oswald, & Best, 2002); MR (Finn, 1982); and ED (Oswald, Coutinho, & Best, 2002). Others, however, have reported an opposite direction of effect, finding that as levels of poverty decrease, minority students are at greater risk for referral as LD (Zhang & Katsiyannis, 2002); MR (Oswald, Coutinho, Best, & Nguyen, 2001); and ED (Oswald et al., 1999).
In order to directly assess the contribution of poverty to the disproportionate representation of African American students in special education, Skiba et al. (2005) studied the relationship of special education enrollment, race, socioeconomic and demographic factors, and test score outcomes in a sample of 295 school districts in a midwestern state. Across ordinary least squares and logistic regression equations, poverty made a weak, inconsistent, and often counter-intuitive contribution to the prediction of disproportionality across a number of disability categories. Where poverty made any contribution above and beyond race in predicting disability identification, its primary effect was to magnify existing racial disparity.
Generalizations about the effects of poverty on parenting may also yield unwarranted assumptions about families from groups overrepresented in special education. Although poverty has been shown to be associated with more negative parenting styles (McLloyd, 1998), there is no evidence that African American or Latino families are, on average, more dysfunctional than other families. Yet, in their recent ethnographic study of racial disproportionality in special education, Harry and Klingner (2006; Harry, Klingner, & Hart, 2005) found negative beliefs about African American families to be pervasive among educators. Families of African American students were described as neglectful, incompetent, and dysfunctional, often absent any firsthand knowledge of those families' actual circumstances. Such descriptions also ignore significant cultural strengths in African American and Latino communities, such as the involvement and expertise of extrafamilial adults, who may act as protective factors despite economic disadvantage (Harry & Klingner, 2006; King, 2005).
In summary, a variety of poverty-associated risk factors have been shown to predict academic and behavioral gaps that might be expected to lead to special education referral, suggesting rhat economic disadvantage makes some contribution to minority disproportionality in special education. Yet the path from initial referral to eligibility determination is complex and governed by policy regulations that are by no means strictly linear. It is not surprising, then, that research to this point has not supported the hypothesis that poverty is the sole or even primary cause of racial and ethnic disparities in special education. In particular, although poverty creates conditions that reduce parenting efficacy, assumptions made about the general quality of African American or Latino families and their contributions to disparate rates of special education referral are unwarranted given the extent of available data.
Finally, regardless of the relationship among poverty, academic achievement, and racial disparities, mechanisms for the negative effects of poverty remain unclear. It is often presumed that economic disadvantage affects educational readiness by increasing biological or family-based risk prior to school entry. Yet students placed at risk for the biological or social effects of poverty are also more likely to attend schools with reduced educational resources and fewer opportunities for quality instruction (McLloyd, 1998; Peske & Haycock, 2006). In an educational system in which poor students of color routinely receive an inferior education, the possible contributions of the schooling itself to disparities in special education service must also be considered.
UNEQUAL OPPORTUNITY IN GENERAL EDUCATION
One of the most consistent findings in educational research is that students achieve in direct proportion to their opportunity to learn (Wang, Haertel, & Walberg, 1997). It might well be expected that students whose educational opportunities are limited will be more likely to be referred for special education services (Artiles & Trent, 1994; Harry, 1994). Differential access to educational resources has been consistently demonstrated for some iminority groups in a number of areas (Kozol, 2005; Peske & Haycock, 2006). Of the possible links between general education practices and special education disproportionality, however, only the proportion of culturally consonant teachers in the teaching force has been directly investigated. Serwatka, Deering, and Grant (1995) found that as the percentage of African American teachers increased, overrepresentation of African American students in the ethotionally disturbed category decreased. Similarly, in a cross-state comparison, Ladner and Hammons (2001) found that the discrepancy of African American and White rates of eligibility for special education rose in direct proportion to the percentage of the teaching force that was White, especially in districts with a White percentage of more than 60%.
More generally, however, inequity in the quality and quantity of educational resources has been extensively documented. Curricula and instructional presentation appear to disfavor working-class students or students of color (Ferri & Connor, 2005, Sleeter & Grant, 1991). Serious deficiencies in physical facilities and resources in urban schools have been documented (Kozol, 1991, 2005; Oakes, Ormseth, Bell, & Camp, 1990). Such resource disparities may have their origin in inequitable school funding formulas (Rebell, 1999) or in historical patterns of segregation and re-segregation (Katznelson, 2005; Orfield & Eaton, 1996). Finally, a number of factors ranging from inadequate teachet preparation (Barton, 2003); to teacher inexperience (Peske &C Haycock, 2006); to teacher reticence to teach in what are perceived to be challenging areas may limit the access of students in high poverty, high minority districts to quality teaching (DarlingHammond, 2004). Students from poverty backgrounds and students of color are also more likely to be taught by teachers with less experience and expertise, in more poorly funded schools that have difficulty recruiting and maintaining both teachers of color in particular and a sufficient teaching force in general (Barton; Donovan & Cross, 2002; Peske & Haycock).
These inequities have a demonstrable effect on the educational opportunity and school achievement of low SES children. In a multiyear observational study, Greenwood, Hart, Walker, and Risley (1994) reported that inferior instruction in low SES schools resulted in students in those schools receiving an equivalent of 57 weeks less academic engagement than students in high SES schools by the sixth grade; as a result, an achievement gap equal to 0.3 of a grade level at school entry grew to a gap of 3.5 grade levels by Grade 6. These data make a strong case rhat students of color in low SES communities ate at greater risk for poor quality educational experiences that undermine their academic achievement.
It is reasonable to presume that factors that limit educational opportunity will impact educational achievement, thereby increasing the risk for special education referral (Skiba, Bush, & Knesting, 2002). Although suggestions that equity in special education services might best be achieved by ensuring that quality educational services for all students are longstanding (Heller et al., 1982), the influence of general educational quality on special education referral is still remarkably understudied. Although the link between teacher demographics and special education disproportionality has been explored to some extent (Ladner & Hammons, 2001; Serwatka et al., 1995), the influence of other systemic factors such as quality of curriculum, instruction, resources, or teacher training on differential rates of special education referral and eligibility determination have yet to be directly studied.
SPECIAL EDUCATION ELIGIBILITY AND DECISION-MAKING PROCESSES
Disparities in special education could be influenced by inadequacies in practice or bias generated at the level of special education referral and decision making. Although this possibility has received some research attention, the pattern of results is somewhat unclear.
Referral. Available data suggest that racial disparities in the classification of students as disabled begin at the stage of initial classroom referral. Reviewing records of students referred for special education evaluation in an urban school system, Gottlieb, Gottlieb, and Trongone (1991) found that teachers referred minority children more often than nonminority children and tended to refer minority students for behavioral rather than academic issues. In a meta-analysis of 10 studies between 1975 and 2000 examining referral to special education, Hosp and Reschly (2003) found that both African American and Latino students were referred more often to special education than White students.
Examination of prereferral decision making by teachers has yielded mixed results. Bahr, Fuchs, Stecker, and Fuchs (1991) found that, despite relatively minor differences in descriptions of academic and behavioral functioning, general education teachers were more likely to describe African American students as difficult to teach and, hence, more likely to be referred to special education. Shinn, Tindal, and Spira (1987) compared teacher recommendations for referral based on curriculum-based measures and found that teachers were more likely to refer Black than White students based on those results in Grades 2 to 4. In contrast, MacMillan and Lopez (1996) found that Black students referred to a student support team prior to special education referral were more likely to have lower test scores and more severe behavioral ratings, leading the researchers to conclude that teachers may wait to refer Black students until their academic or behavioral problems reach a higher level of severity. On a positive note, Gravois and Rosenfield (2006) found that changes in prereferral practice can significantly impact disproportionate representation: Schools using an instructional consultation model significantly reduced both their overall rate of special education referral and identification and reduced racial/ ethnic discrepancies in rates of referral and identification.
Assessment and Decision Making. Investigations of the possibility of bias during the assessment and decision-making process have not been undertaken recently and present a somewhat conflicting picture. Analogue studies using a case study vignette (e.g., Priero & Zucker, 1981) found a greater willingness among both general and special education teachers to recommend minority students for special education given identical referral information. In two studies using a similar simulated research paradigm, Tobias and colleagues (Tobias, Cole, Zibrin, & Bodlakova, 1982; Tobias, Zibrin, & Menell, 1983) found that teachers rated students of minority backgrounds different than their own as more appropriate for special education identification in the first but not the second study. Reviewing tapes of case review teams making placement decisions, Ysseldyke, Algozzine, Richey, and Graden (1982) reported that factors such as student race and SES contributed more to placement decisions than did performance data. Tomlinson, Acker, Canter, and Lindborg (1978) examined special education referral and decision-making processes and found that minority students were referred more often, that their parents were contacted significantly less often to patticipate in the special education process, and that the recommendations to minority parents were more restrictive and less comprehensive than recommendations for nonminority parents.
Large discrepancies between actual practice and the ideal due process provisions outlined in IDEA have been documented in the literature, and those discrepancies may well contribute, to some extent, to disproportionality in service. Gottlieb, Alter, Gottlieb, and Wishner (1994) noted that, in the urban school districts they studied, many students received services for learning disabilities despite not meeting the LD discrepancy criteria for identification. Similarly, MacMillan and Reschly (1998) argued that up to half of all students identified as LD do not meet their state's criteria for identification. In their ethnographic exploration, Harry and Klingner (2006) described numerous inconsistencies in the special education conferencing phase that may contribute to disproportionality, including rates of special education referral differing by the race and ethnicity of the teacher, the disproportionate weight given the opinion of the referring teacher at the case conference, and the weak emphasis on prereferral strategies.
Thus, racial and ethnic disparities in special education identification appear to begin at the stage of initial teacher referral, and it seems likely that breakdowns in the due process provisions governing special education can contribute to the inappropriate identification of minority students in special education. Yet given the lack of consistency in this research, as well as the age of many of the studies, the extent to which current special education eligibility determination processes contribute to special education inequity is unclear. The most recent National Research Council (NRC) panel (Donovan & Cross, 2002) concluded that evidence of bias in the referral to placement process was mixed, but that the process has sufficient shortcomings as to be unable to ensure that the correct students are being identified. Further, the panel contended that the entire process is weighted toward referral and placement only after a student has experienced failure, thus ensuring that child's problems will be relatively intractable by the time he or she is finally placed in special education. BEHAVIOR AS THE NEXUS OF RACE AND DISABILITY
Special education is, of course, not the only educational domain in which students of color are disproportionately represented. Consistent evidence has documented large gaps between students of color and their peers in academic achievement as measured by accountability test scores (Jencks & Phillips, 1998); graduation and dropout rates (Holzman, 2004); and placement in educational programs such as gifted and talented and Advanced Placement/Honors courses (Donovan & Cross, 2002; Ford, Grantham, & Whiting, 2008; Joseph & Ford, 2006).
The disproportionate iepresentation of African American students in school suspension has been widely documented. For more than 30 years, in national, state, district, and local data, African American students have consistently been found to be suspended out- of-school at higher rates than other students, and similarly overrepresented in office referrals, corporal punishment, and school expulsion (e.g., Children's Defense Fund, 1975; Raffaele Mendez & Knoff, 2003; Skiba, Michael, Nardo, & Peterson, 2002; Wu, Pink, Crain, & Moles, 1982). In one study of a large and diverse school district, 50% of African American male and one third of African American female middle school students experienced out-ofschool suspensions during one school year (Raffaele Mendez & Knoff), rates that were substantially higher than White male (25%) and White female (9.3%) middle school students. Disproportionality in school suspension has not been as consistently documented for Latino or other ethnic minority groups (Skiba & Rausch, 2006).
The contributing factots or causes of racial and ethnic disparities in school discipline have not been conclusively determined. Although it has been argued that disproportionality in school punishments is primarily a function of poverty (National Association of secondary School Principals, 2000), race remains a significant predictor of suspension and expulsion, even when socioeconomic status is controlled in multivariate analyses (Skiba, Michael, et al., 2002; Wu et al., 1982). Nor does disciplinary disproportionality appear to be the result of differential rates of misbehavior by African American students. Any racial differences in reasons for suspension that have been found suggest that African American students receive more severe punishments for less serious infractions (Shaw & Braden, 1990) or are referred to the office more frequently for more subjective reasons, such as disrespect or loitering (Skiba, Peterson, et al., 2002). Other explanations for disciplinary disproportionality include the possible misinterpretation by classroom teachers of culturally based behaviors (Townsend, 2000) or stereotypes regarding Black males that increase the likelihood of office referral (Ferguson, 2001).
There are also indications of racial disproportionality in the application of the specific disciplinary provisions of IDEA. A recent state report (Rausch & Skiba, 2006) found that about 3% of African American students with disabilities received at least one of the IDEA disciplinary provisions, a rate 2.8 times higher than all other students with disabilities. Futther, the greatest racial disparities were found in the IDEA disciplinary provision other suspension/expulsion greater than 10 days, in which African American students were found to be 3.4 times as likely as their peers with a disability to receive this provision. Disproportionality in specific school districts ranged from relatively proportional use (relative risk ratio = 1.03) to a rate in one school district in which African American students with disabilities were more than 10 times more likely than other students with disabilities to receive one of the IDEA disciplinary provisions.
The intersection of disproportionality in school discipline and special education has been commented on (Gregory, 1997) but insufficiently explored. Investigations of disproportionality in referrals to special education or prereferral teams consistently find that African American students are more likely to be referred for behavioral reasons (Gottlieb et al., 1991; MacMillan & Lopez, 1996). The nature and causes of disciplinary disproportionality represent an important avenue for further research on racial disparities in special education.
CULTURAL MISMATCH AND CULTURAL REPRODUCTION
Emerging scholarship has conceptualized the disproportionate representation of minority students in special education, African American students in particular, as a symptom of a broader disconnect between mainstream educational culture and the cultural orientations of communities of color. A number of scholars have argued that contemporary mainstream educational systems, special education systems in particular (Patton, 1998), closely reflect the knowledge, values, interests, and cultural orientations of White, middle-class cultural groups (Delpit, 1995; King, 2005). Education that fails to explicitly teach the codes and rules necessary for successful participation in unfamiliar cultural contexts (Delpit), does not connect knowledge produced in schools to students' lived experiences (Ladson-Billings, 1994), or ignores the foundational role of culture in knowledge production (Sheets, 2005) may yield inadequate and inappropriate educational experiences for a range of cultural groups.
Notably, such knowledge is not well-represented in mainstream scholarship (Trent et al., 2008). The intensive observation required by such research may make it more difficult to conduct, compared to tests of more prevalent hypotheses present in contemporary scholarship (e.g., poverty, test bias). Alternatively, it has been argued that non-mainstream epistemologies, paradigms, discourses, and research orientations have been systematically devalued or "silenced" (Delpit, 1995), producing a database that has explored only a limited range of hypotheses for unequal educational outcomes of African American and Latino students in general (King, 2005), and disproportionality in special education in particular (Patton, 1998).
One theoretical perspective that holds promise for providing a framework within which to view racially disparate educational outcomes is the model of cultural reproductive systems and actions (Bowles & Gintis, 1976). Developed as an explanation of the perpetuation of social class hierarchies, the theoretical framework of cultural reproduction has been utilized by equity researchers to demonstrate how institutional and individual actions maintain a hierarchical status quo at the expense of less-privileged groups (Harry & Klingner, 2006; Mehan, 1992; Oakes, 1982). Cultural reproduction implies that individuals can become a part of institutional patterns through constitutive actions (Mehan, 1992; Mehan, Hertweck, & Miehls, 1986) that can reproduce the status quo without being consciously aware of their conttibution to inequity.
Recent ethnographic invesrigations have found clear evidence of reproductive processes that may well contribute to inequitable outcomes in special education. In an ethnographic study focusing primarily on the role of school psychologists in assessment decision making, Harry et al. (2002) found that although psychological testing is often perceived as an objective procedure designed to reduce the influence of individual judgment, in fact, the process is often highly idiosyncratic, as psychologists choose tests or test batteries more likely to produce the results they, or the teachers making the referral, wish to see. Using Heller et al.'s (1982) conclusion that disproportionality could be viewed as a problem if there is evidence of inappropriate practice or bias at any phase of the process, Harry and Klingner (2006) tracked opportunity to learn, the special education eligibility decision-making process, and special education programming. They found evidence of a number of institutional constraints and constitutive actions that appeared to influence the course of special education placement and programming for minority students, including poor teacher quality, large class sizes, arbitrary application of eligibility decision-making criteria, tardiness in placement processes, and special education programs that were themselves ineffective or overly restrictive. The authors argued that such findings suggest the need for increased attention to school-based risk as a contributing factor to inequity in special education.
DISPROPORTIONALITY AS A MULTIPLY DETERMINED PHENOMENON
It should be apparent from the preceding discussion that there is no single simple explanation that appears to fit the data on special education disproportionality. Rather, minority disproportionality in special education appears to be multiply determined, a product of a number of social forces interacting in the lives of children and the schools that serve them (see Trent et al., 2008).
Qualitative findings have highlighted the interacting forces that may set the context for and maintain racial disparities in special education. In an intensive case study interviewing teachers, principals, school psychologists, and administrators about their perspectives on special education and culture, Skiba et al. (2006a) reported a complex picture of the factors that contribute to referral. Teachers feel highly challenged to meet the needs of students with economic disadvantages, yet feel they are given insufficient resources to meet those needs. Classroom behavior proved to be a difficult issue for many teachers, exacerbated by cultural gaps and misunderstandings. Prereferral or general education intervention teams were seen as potentially useful in supporting teachers working with students with academic or behavioral challenges, but the use and perceived effectiveness of those teams varied widely. Perceiving special education as the only resource available for helping students who are not succeeding, classroom teachers were quite willing to err in the direction of over-referral if it meant access to more resources for their neediest students. Finally, there was clear discomfort among many respondents in discussing issues of race; although comfortable and even eloquent in describing the impact of poverty, many respondents seemed anxious to avoid talking about issues involving race or ethnicity. The multidetermined nature of disproportionality likely means that there is no single cause that can be called on to explain racial and ethnic disparities in special education in all states or school districts. In urban schools and districts, a lack of physical and personnel resources may create a pressure to refer low performing students who are predominantly minority to one of the few services available for students who are struggling (Gottlieb et al., 1994; Skiba et al., 2006a). Yet Ladner and Hammons (2001) found that the highest rates of racial/ethnic disparities in special education service were not evident in those urban districts, but rather in higher-income suburban districts. These types of discrepancies suggest that the search for the causes of disproportionality will need to become more attuned to differential rates of disproportionality across locales and different factors that may contribute to disproportionality in those locales. Widely differing racial/ethnic patterns of disproportionate representation suggest that the causes of disparities will vary considerably for African Americans in Washington, DC or New York City and Latino students in Houston or Los Angeles, and that both of these will show a pattern of disproportionality dramatically different from a predominantly White school system in a suburban or rural location.
STRATEGIES FOR REDUCING DISPROPORTIONATE REPRESENTATION
If disproportionality in special education is multiply determined, no single intervention strategy can be universally relied on to reduce racial disparity. Rather, complex causality clearly suggests the need for comprehensive and multifaceted assessment and intervention plans. In particular, the possibility that the determinants of disproportionality are locale-specific suggests that remediation plans must be driven by local needs assessmenr capable of identifying unique local patterns. Team-based needs assessment models for addressing disproportionate representation have been described by Ritter and Skiba (2006) and Klingner et al. (2005). Central to such an approach is a process that moves from data collection and examination, to interpretation, to culturally competent intervention and evaluation.
EXAMINATION OF THE DATA
Data on disproportionality selves to establish both a baseline and a method of monitoring progress. The NRC recommended a national effort to establish both a standard data collection system and a longitudinal assessment of trends in disproportionality (Donavan & Cross, 2002). One important future course for practical remediation of disproportionality at the local level will be to disseminate practical methods of data collection analysis (Salend, Garrick Duhaney, & Montgomery, 2002). It seems likely that the continuous feedback loop afforded by the examination of local data on racial disparities can create change at the systems level (Johnson, 2002). Yet, it is only relatively recently that the field of special education has identified a set of measures (e.g., the composition index, the relative risk ratio) with which to monitor disproportionate representation. If local efforts are to be made to address racial disparities, practical and efficient methods for calculating disparities will need to become available to school personnel.
A range of possible hypotheses might be brought to bear in interpreting a set of data indicating racial disparity. On one end, hereditarian interpretations (e.g., Herrnstein & Murray, 1994) have tended to focus on inherent and genetic explanations of the achievement gap and group differences in performance. Alternatively, critical race theory (Delgado & Stefancic, 2002; LadsonBillings & Tate, 1995) suggests that racial and economic disparities result from the use of the concept of race in structuring institutions and interactions to maintain the power and privilege of the dominant group. It is clear that each of these theoretical orientations yields very different implications for intervention.
Indeed, the effectiveness of an intervention chosen to address disproportionate representation depends, to some degree, on the accuracy of diagnosis of the causes of disparity. Early intervention appears to be an extremely promising intervention for a range of developmental issues related to socioeconomic disadvantage (Barnett, 1995). Early intervention approaches could be expected to reduce disparities only to the extent that economic disadvantage is at work. Early intervention would not be expected to address systemic failures or bias and would hence fail to address disproportionality that is due to institutional inequity.
Unfortunately, interpretation of data on differential racial treatment itself appears to be conditioned by race. The difficulty that educators, especially White educators, have in openly talking about race and racism has been well documented (King, 1991; Skiba et al., 2006a; Trepagnier, 2006). A number of authors have noted that it is common for interpretations of equity data to be based on a majority viewpoint (King, 2005; Patton, 1998). Recent history from the Simpson trial to reactions to Hurricane Katrina indicate that, at this point in our nation's history, interpretations of data on racial and ethnic disparities will vary depending on the cultural makeup of the audience confronting the data. Thus, educators and policy makers seeking effective interventions to close special education equity gaps must be willing to openly discuss and address issues of race, ethnicity, gender, class, culture, and language. Moreover, processes chosen to address inequity must have at their core a mechanism to ensure that the perspectives of all stakeholders, especially those of historically marginalized groups who have been