Intellectual, academic, and behavioral functioning of students with high-incidence disabilities: a cross-categorical meta-analysis.
A meta-analysis of 58 studies was performed in which IQ, academic achievement, and behavior characteristics were examined across students with learning disabilities (LD), mild intellectual disabilities (MID), and emotional/behavioral disabilities (E/BD). The effect sizes between students with LD and MID were the largest in the domains of IQ and academic achievement, and the differences involving pupils with E/BD versus those with LD and MID were largest in the behavior realm. Very little disparity was found when comparing (a) those with LD and students with E/BD on IQ measures, and (b) pupils with LD and those with MID in behavior. Implications for cross-categorical and noncategorical special education are discussed.

Behavior evolution (Research)
Disabled students (Research)
Disabled students (Education)
Intellect (Research)
Intelligence levels (Research)
Academic achievement (Research)
Sabornie, Edward J.
Cullinan, Douglas
Osborne, Susan S.
Brock, Lynne B.
Pub Date:
Name: Exceptional Children Publisher: Council for Exceptional Children Audience: Academic; Professional Format: Magazine/Journal Subject: Education; Family and marriage Copyright: COPYRIGHT 2005 Council for Exceptional Children ISSN: 0014-4029
Date: Fall, 2005 Source Volume: 72 Source Issue: 1
Event Code: 310 Science & research Canadian Subject Form: Behavioural evolution
Geographic Scope: United States Geographic Code: 1USA United States
Accession Number:
Full Text:
It has been well over 25 years since Hallahan and Kauffman (1977) published their seminal piece in The Journal of Special Education entitled, "Labels, Categories, Behaviors: ED, LD, and EMR Reconsidered." In that article the authors concluded that little evidence existed to justify differential special education treatment in the three categories of high-incidence disability--learning disability (LD), mild intellectual disability (MID), and emotional/behavioral disability (E/BD). The authors stated, "It is nearly impossible to separate children meaningfully into these three areas of special education" (p. 144). As a result of opinions typical of those found in Hallahan and Kauffman and others in support of noncategorical special education service delivery (see Algozzine & Ysseldyke, 1983; Marston, 1987; Reschly, Tilly, & Grimes, 1999; Reynolds & Birch, 1977; Ysseldyke & Marston, 1999), and because of public school administrative convenience and cost, many students with the three types of high-incidence disabilities are now educated side-by-side in the same educational environments.

Although Hallahan and Kauffman's (1977) conclusions were an early catalyst for cross-categorical special education service delivery in the schools, more recent opinions on this issue extend far beyond those first offered in the late 1970s. The issue of cross-categorical or noncategorical special education appears far from dead and now more complex than when first brought to light. Ysseldyke and Marston (1999), for example, concluded that categorical special education in the schools is problematic, that it can no longer be justified, and that "diagnostic efforts to differentiate groups should be diminished" (p. 6). Reschly and Tilly (1999) called for elimination of the present categories in special education, especially those related to students with high-incidence disabilities, and use of a noncategorical system with different functional criteria used for eligibility. Tilly, Reschly, and Grimes (1999) recommended a "problem-solving assessment" model for noncategorical special education identification and eligibility, as well as program planning. It is clear that the noncategorical service delivery model has also gained considerable Support. Careful analysis of studies comparing two, or all three, groups should lead to better understanding of all individuals identified as having high-incidence disabilities.

Perhaps the most influential work that may determine whether educators think students with high-incidence disabilities are truly similar or dissimilar for instructional purposes does not originate in research. The President's Commission on Excellence in Special Education (PCESE; 2002), in an effort to streamline identification and eligibility for special education purposes, suggested that the term "developmental disabilities" should include "specific learning disabilities (SLD), speech and language impairments, emotional disturbance, mild mental retardation and developmental delay" (p. 21). The experts on the PCESE did not see the need for continuing the 13 categories of disabilities found in earlier versions of the Individuals with Disabilities Education Act (IDEA). The Consortium for Citizens With Disabilities (CCD; 2002), a coalition of nearly 100 disability organizations, challenged the conclusions of the PCESE and found it inappropriate to group the various disability categories together. The CCD promised to "fiercely resist" the alteration of the 13 disability categories under the IDEA and stated that the essence of the issue should be whether the current system correctly identifies all who are eligible. Time will tell whether the recommendations of the PCESE are the best way to categorize those with high-incidence disabilities for schooling. One necessity to solve the categorization issue is a comprehensive analysis of research that examined similarities and differences that exist among students in the categories of high-incidence disabilities.

Since 1977 there have been more than 150 published articles, studies, monographs, and book chapters that discussed similarities and differences across students with the three types of high-incidence disabilities. A synthesis of the high-incidence disabilities cross-categorical research, however, does not exist. This synthesis could help defend or refute the arguments concerning the need for categorical, cross-categorical, or noncategorical special education in the schools. Because of the plethora of studies that are now available that have examined similarities and differences across the three types of high-incidence disabilities, and because of the support that cross-categorical and noncategorical special education service delivery possess, it is now time to reexamine the accuracy of previous conclusions regarding the similarity of students across the three categories of high-incidence disabilities.

When Hallahan and Kauffman (1977) compared and contrasted the cognitive and behavioral characteristics of students with LD, MID, and E/BD, they did not have access to meta-analysis or the ability to cluster and synthesize quantitative results of research. Meta-analysis uses study means, standard deviations, and sample size (and other data) in quantitative research to explain the magnitude of difference between two groups on the same dependent variable. An effect size (ES) is derived in meta-analysis that is an estimate of the difference, or lack thereof, between two samples using their respective means, standard deviations, and sample size. A major advantage of the ES statistic is that it is a common metric that permits the combining of results from many studies on the same construct. The ES aggregation in a meta-analysis allows for an indication of an overall pattern or strength of difference between two groups. A meta-analysis not only takes into consideration study means and standard deviations forming ESs but also is dependent on sample sizes. Studies with larger samples attain more weight and contribute more robustly to the overall results of a meta-analysis than do investigations with smaller samples. In other words, a study yielding an ES of .30 with a sample of 500 participants will Contribute more to the overall meta-analysis results than the same .30 ES with a sample of 50. Meta-analysis shows that the larger the ES--in either a positive or negative direction--the better the distinction between two groups.

The purpose of the meta-analysis found herein was to determine whether it is now possible to meaningfully separate students into the categories of LD, MID, and E/BD on the basis of research. Our meta-analysis was limited to three dependent variables that are important and necessary for special education eligibility and that make a difference when trying to educate students with disabilities. We combed the literature to find studies to be grouped into the following domains: IQ, academic achievement, and behavior. The meta-analysis was limited to these dependent variables for they define in large part how students perform in school. Statements concerning the three variables of interest are found in the definitions of each category of high-incidence disability, and performance on the three dimensions accounts for a majority of students identified as disabled and eligible to receive special education (Reschly & Tilly, 1999). Without consideration of IQ, academic achievement, and behavior it is nearly impossible to describe objectively any student with a LD, MID, or E/BD.

The three domains of interest chosen herein also accurately describe the who in special education (Bateman, 1994). The who, along with the how and where, are core concepts and perpetual questions of great importance in the field of special education (Kauffman & Landrum, 2006). Our modest, but worthy goal herein was to provide greater illumination of the who in the categories of LD, MID, and E/BD. We attempted to answer the following global research question: Do meaningful, educationally related and quantitative differences exist across school-identified students in the three categories of high-incidence disabilities?


Studies that compared two or three of the high-incidence disability categories were located in the Educational Resources Information Center (ERIC), EdInfo, Education Abstracts, and PsycInfo databases between the years 1977 and 2003. Different keywords were used for the categories of E/BD (e.g., emotional disturbance, behavior disorder, etc.) and MID (e.g., educable mental retardation, mild cognitive disability, developmental disability, etc.) because of the variety of titles used in different states and studies. In addition, a hand search, or "footnote chasing" was performed on the reference lists of the articles published during the 1980s, 1990s, and early 2000s so that additional studies could be found and examined for possible inclusion. This initial literature search yielded a total of 152 reports that included chapters, discussion pieces, articles in monographs, position papers, dissertation abstracts, and empirical studies that could be used for our preliminary research inspection for the meta-analysis.


We examined each research report to refine its suitability for further analysis and devised specific a priori criteria for inclusion in the present meta-analysis. The inclusion criteria were selected after discussion among the authors and to partially control for study quality following the suggestions of Lipsey and Wilson (2001). Our study inclusion rationale included the following:

1. Samples of two, or all three types of high-incidence disabilities had to be represented. Participants had to have been previously recognized (i.e., before the research was conducted) as having a school-identified high-incidence disability, and not because of the actual assessment performed in the research examined for the meta-analysis.

2. Participants had to be identified as having a high-incidence disability under IDEA and corresponding state law active at the time the study was published, and only in the United States.

3. Each study had to use norm-referenced, standardized tests of achievement, ability, and behavior and report the results in a traditional manner. Each test used in the studies had to have known reliability and validity statistics. Because assessment for special education eligibility depends heavily on norm-referenced test performance, and because such measures are also used to judge students' progress in school over time, we wanted to use traditional data that are typically found where students with high-incidence disabilities are educated.

4. The investigation had to report sufficient statistics (i.e., means, standard deviations, sample sizes, t-test and F-ratio results, etc.) for the groups of interest on at least one of the three dependent variables.

5. The samples had to be of proper age covered under IDEA. In light of our specific interest in students eligible for special education in the United States under IDEA, we limited the research samples in studies to only those between the of ages of 3 to 21, or that included data on students when they were between those ages.

6. Studies found in the behavior domain had to have either parent or teacher (or both) ratings of participants' conduct or behavioral traits. We did not allow for participant self-ratings of behavior or personality characteristics.

7. Each study had to be a group design investigation. In keeping with the recommendations of Lipsey and Wilson (2001), we did not mix single-subject designs with the group design studies sought for the present analysis.

8. Each study had to be published in a peer-reviewed journal or similar outlet (e.g., peer-reviewed monograph).

This selection system reduced the original total of 152 related literature items to 58 service able research reports from 26 different peer-reviewed journals. Our selection criteria reduced the final number of published literature to just more than one third of that originally found; most of the peer-reviewed studies excluded from the present meta-analysis were used in an additional descriptive review of the literature (see Sabornie, Evans, & Cullinan, 2006).

Two of the co-authors completed the literature search, and there was 100% agreement on matters of study inclusion, data to be used for computer analysis, variables, and samples of interest. Coding of study data was also performed by the same two co-authors. A reliability check of the coders' study data was performed on 20 (34%) randomly selected studies used in the meta-analysis. This intercoder agreement check was used to verify that statistics used in the effect size computations (using computer programs; see "Effect Sizes" in the following) were exactly as they appeared in the original published study involving sample sizes, means, standard deviations, t-test, and F-ratio data. This check involved a total of 288 separate comparisons. The intercoder agreement percentage was calculated by summing the agreements and then dividing the sum by the total number of agreements plus disagreements, times 100. Two minor errors (decimal place values) were found during the intercoder agreement calculations for an overall concurrence rate of 99.3%.


The reporting of detailed participant demographics varied widely across the available studies used in the present analysis. The participants in the 58 studies used for the present meta-analysis totaled 14,528 individuals with high-incidence disabilities. Those with LD numbered 7,876 (54%), there were 3,057 (21%) persons with MID, and 3,595 (25%) participants were identified as having E/BD. Twenty-three (40%) of the selected studies reported only on preschool (n = 1) or elementary level participants in the categories of high-incidence disabilities, 14 (24%) of the studies examined a combination of elementary and middle school level students, 6 (10%) studies used only middle school age pupils, 2 (3%) studies used a combination of middle school and high school age participants, 6 (10%) studies inspected only high school aged students, and 7 (12%) studies used participants from the elementary to high school levels (percentages were rounded).

Only 28 (48%) studies examined herein provided separate gender data related to all the participants. In these studies there were 4,500 male participants (71%) and 1,876 female participants (29%). This represents 44% of all the participants examined in the present analysis. Among the studies reporting gender statistics of the samples, of the pupils with LD 71% were boys, of students with MID 56% were boys, and in the group with E/BD 78% were boys. Very few studies reported IQ, academic achievement, or behavior scores separately by gender in cross-categorical fashion.

Twenty-two (38%) of the 58 studies in the analysis reported racial demographics that could be separated into disability designation. The 22 studies with useable racial data included 28% of the total participants in this report. Sixty-five percent of the reported participants were Caucasian, 31% were African American, 2% were Latino, and 2.5% "other." In some cases the other category included all participants who were non-Caucasian. Again, few studies reported participants' performance separated by race. It would be unwise to draw any conclusions from the gender or racial data found in this analysis because of the low number of studies that reported participants' characteristics in these areas.


Using the computer statistical packages JMP (SAS Institute, 2001) and ES (Shadish, Robinson, & Lu, 1999), study data were coded into the three domains of interest, groups examined, and study and sample characteristics. Data were also coded into outcome measures and statistics necessary for calculating effect sizes and related information including pooled standard deviation (SD) per study, standard error (SE) of the ES, inverse variance weight of the ES (IVW), IVW x ES, weighted mean ES ([ES.sub.w]), SE of the [ES.sub.w], 95% confidence intervals around the [ES.sub.w], and Q statistic. The formulas for all of the above ES and related statistics originate with Lipsey and Wilson (2001) and are found in the Appendix. For studies with means, SDs, t-test results, and F ratios reported, ESs were calculated for each two-group comparison (i.e., LD vs. MID, MID vs. E/BD, and LD vs. E/BD) on each dependent variable measure using Cohen's d statistic, or the standardized mean difference (SAID) methodology. The SMD calculation with pooled SDs used in the formula denominator (see Appendix) was chosen because in each study the participants with LD, MID, and E/BD were already assigned group membership and were nonexperimentally defined (i.e., there were no randomly assigned or selected treatment vs. comparison group investigations). Using a two group, pooled SD in ES calculations that do not include traditional experimental and control groups has been used previously and is the recommended procedure (Lipsey & Wilson; Rosenthal, 1991).

Many studies provided the opportunity to obtain more than one d statistic per investigation. In such cases we followed the recommendation of Lipsey and Wilson (2001) and calculated separate Cohen's d statistics and separate pooled SDs, one for each comparison of interest. We then averaged the d statistics and pooled SDs to calculate one single ES and pooled SD for each study. This was performed in the IQ (e.g., averaging ESs for verbal and performance IQs), academic achievement (e.g., averaging ESs for academic achievement subtest scores), and behavior (e.g., averaging ESs for behavior subtest scores) domains. As a check for this data aggregation method, in each of the two group comparisons across high-incidence disabilities (e.g., LD vs. MID, LD vs. E/BD, MID vs. E/BD) and three dependent variable domains, the Pearson product-moment correlations between the number of d statistics produced within a study and average ES per study were not significant (at the p < .05 level). In other words, averaging the ES data within an investigation to produce one ES per study did not overestimate group differences. Homogeneity analyses (see the following) also verified that our chosen method of study data and ES aggregation were appropriate.

A data coding consideration in the behavior domain was the direction of the ES based on the characteristics of the instrument used to rate participants' traits. Studies that used the Behavior Problem Checklist, on which lower scores equal more appropriate behavior, had to be coded differently from studies using other tests (e.g., SSRS--social skills subtest) where higher scores translate into positive outcomes (K. A. Kavale, personal communication, June 21, 2000). The ES (Shadish et al., 1999) program allows for such data entry nuances, and we used it exclusively in such situations.

Cohen's d statistic, reflecting the magnitude of group difference, was derived for each available dependent variable in a study and for each available two-group comparison. We were unable to compile d statistics using sample means and standard deviations in four studies. In Scruggs and Mastropieri (1986), for example, two ESs using t-tests for second- and third-grade students with LD versus pupils with E/BD (in academic achievement) were provided in the study. In this case we averaged the two ESs provided into one study ES for Scruggs and Mastropieri. In two additional studies (Harris, King, Reifler, & Rosenberg, 1984; Sprafkin & Gadow, 1987), the available data allowed for using the ES program (Shadish et al., 1999) to calculate Cohen's d with sample size and t-test results. Gajar (1979) provided sufficient data to use the ES program to compute Cohen's d using sample sizes with F ratios found in the report.

Homogeneity Testing. If a study had dependent variables in both IQ and academic achievement (e.g., Kortering & Braziel, 1999), the IQ scores were placed in the IQ ES domain, and the academic achievement scores were used in an ES calculation for that domain. If a study investigated dependent variable performance for all three groups (e.g., Cullinan & Epstein, 1985), then that study produced three study effect sizes. Thus, a total of 130 ES measurements were derived from the 58 studies for an average of 2.2 ESs per study.

An important aggregation test used in meta-analyses is that of homogeneity, which attempts to determine whether studies can be described as sharing a common effect size or population ES (Hedges & Olkin, 1985; Lipsey & Wilson, 2001). Because a major apprehension in any meta-analysis is the extent to which studies are commingled to find a common ES for two groups on a dependent variable, testing for homogeneity is a characteristic of any good meta-analysis (Davies & Crombie, 2003). The homogeneity test is based on the Q statistic. We performed a test of homogeneity to yield a Q statistic in each dependent variable domain and for each two group comparison using the formulas found in the Appendix. The Q statistic, when summed across all studies in a domain, is distributed as the [chi square] statistic with n - 1 degrees of freedom. A non-significant [chi square] statistic indicates a homogeneous distribution. "In a homogeneous distribution, the dispersion of the ESs around their mean is no greater than that expected from sampling error alone (the sampling error associated with the subject samples upon which the individual effects sizes are based). In other words, in a homogeneous distribution an individual effect size differs from the population mean only by sampling error" (Lipsey & Wilson, p. 115). A nonsignificant test for homogeneity (or Q statistic), therefore, produces statistical support for the ES aggregation method used in a meta-analysis. The pooling of ES method used herein (i.e., one study, contributing one domain ES, based on averaging all within-study ESs on the same construct) was shown to be proper for all group comparisons across all three domains. Each two-group comparison in each domain was shown to be homogeneous, and the representative Q statistics are presented in Table 1. Given the favorable Q statistic results, therefore, and using best practice in meta-analysis (i.e., according to Cooper & Hedges, 1994), no further analysis of our aggregation method was necessary.


The outcome variables examined in the IQ domain included full-scale, verbal, and performance IQ scores on norm-referenced intelligence tests such as the Wechsler scales (all versions except for the WPPSI; i.e., WISC, WISC-R, WISC-III, WAIS), Kaufman Assessment Battery for Children (K-ABC), Stanford-Binet (all versions), and McCarthy Scale General Cognitive Index. Several studies reported full-scale IQ measures and separate results for verbal and performance IQ scores. In such cases we used only full-scale IQ scores. We also included IQ data from studies where participants were examined in domains other than IQ (e.g., Sabornie, Kauffman, & Cullinan, 1990), but presented group means and SDs on IQ in matching of participants on various demographic variables. For the IQ dependent variable, 11 studies examined all three groups with high-incidence disabilities, 14 studies compared IQ across only the groups with LD and MID, and 15 separate studies investigated IQ differences with only those having LD versus participants with E/BD.

The outcome variables used in the academic achievement domain were scores from norm-referenced academic achievement tests such as the Woodcock-Johnson Psychoeducational Battery (all versions), Wide Range Achievement Test (all versions), the achievement scale of the K-ABC, Kaufman Test of Educational Achievement, Peabody Individual Achievement Test, Clinical Evaluation of Language Functions, and Gilmore Oral Reading Test. With the exception of three studies that reported grade level performance, the majority of studies used in the academic achievement domain reported standard scores of participants' performance. In the academic achievement domain, six studies compared all three groups with high-incidence disabilities, three contrasted performance of only those with LD versus participants with MID, and only nine studies examined differences with students identified as either LD or E/BD.

The outcome measures examined in the behavior domain included scores on norm-referenced instruments that assessed various behavioral traits of the participants. Tests included the Behavioral and Emotional Rating Scale, Behavior Problem Checklist, Social Skills Rating System, Children's Adaptive Behavior Scale, Social Behavior Assessment-Teacher Version, Conners Teacher Rating Scale (and abbreviated version), Homework Problem Checklist, Walker-McConnell Scale of Social Competence and School Adjustment, School Social Behavior Scales, Child Behavior Checklist, Attention Deficit Disorders Evaluation Scale, and Teacher Ratings of Social Skills. Behavioral ratings of the participants were provided by both teachers and parents and included behavioral and emotional strengths, general overt behavior and problem behaviors, social competence and other socially oriented behavior, depression-like symptoms, perceived academic behavior, and hyperactivity problems. Last, in the behavior domain, nine studies compared behavioral traits across all three groups of students with high-incidence disabilities, two research reports uncovered differences in participants only with LD or MID, and nine investigations examined students with LD and E/BD only.



We performed Pearson product-moment analyses between the year of study publication and the ES for each study across all dependent variable domains and two group comparisons. The only significant correlation (at the p < .05 level) found was between year of publication of the study and its corresponding within-study ES for groups with MID and E/BD in IQ (r = -.63, p < .05). As year of publication increased, the IQ ES between these two groups decreased, or vice versa. The mean year of publication of the 58 studies was 1990 (median = 1989).

Correlational analyses were also performed on the age of participants in each study with calculated within-study ESs across domains and two group comparisons. The only significant correlation in these analyses concerned the comparison between students with LD versus those with MID in the behavior domain. The age of participants in these two groups was significantly negatively correlated (r = -.71, p < .05) with the within-study ES. As the age of these two types of participants increased, the within-study ES decreased, or vice versa.


A weighted mean ES ([[bar.ES].sub.w]) was calculated for each of the two group comparisons in each of the three domains of interest. As in all meta-analysis calculations, a positive [[bar.ES].sub.w] favors the first group in the comparison and a negative [[bar.ES].sub.w] favors the second group. Table 1 presents the major findings for the meta-analysis performed herein.

IQ Domain. Little IQ difference can be found in the studies comparing performance across the groups identified as LD and E/BD. In the remaining two IQ comparisons involving participants with MID, however, statistically significant weighted mean ESs separate such students with those having either LD or E/BD. Because [[bar.ES].sub.w] estimates can be translated similar to z scores (i.e., an ES of 1.00 = 1 SD), the [[bar.ES].sub.w] of [+ or -] 2.17 or greater means that more than 98% of students with either LD or E/BD outscored the average student with MID.

Academic Achievement Domain. The [[bar.ES].sub.w] estimates across studies in the academic achievement domain span the range from weak to strong differences across the three comparison groups, but each one was statistically significant. The standardized test performance levels between students with MID versus those with E/BD showed the greatest disparity, and the comparisons between participants with LD and pupils with E/BD showed the least. The moderate levels of academic achievement difference between students with LD and MID indicate that at least two thirds of the group with LD outperformed the average student with MID. Approximately 76% of students with E/BD scored higher than the typical pupil with MID.

Behavior Domain. The [[bar.ES].sub.w] estimates in the last domain of interest provide somewhat surprising differences across the comparison groups. What is particularly striking is the lack of difference in many types of school-related behavior between students with LD and those with MID. It is not surprising, however, that students with LD and pupils with MID both significantly outperform participants with E/BD in teacher- and parent-rated behavior. Approximately 76% of students with LD were rated better than the average pupil with E/BD, and about 77% of youth with MID surpassed the average person with E/BD. The finding concerning the statistically significant difference between youth with MID and those with E/BD is surprising given the fact that students in the former group have their own set of adaptive behavior problems. Figure 1 presents a schematic of the various weighted mean effect sizes across groups and domains.



This study compared and contrasted some of the school-related characteristics of students with LD, MID, and E/BD. The results show that students with high-incidence disabilities can be separated categorically on measures of IQ, academic achievement, and behavior.

One group that stands out remarkably from the other two in the meta-analysis is the cluster with MID. The group with MID differs significantly from groups with other high-incidence disabilities, especially in the IQ and academic achievement domains. Given the parameters of the federal definition of intellectual disability, the cognitive domain differences demonstrated here for persons with MID are not surprising. Conversely, it is possible that students with MID performed more poorly in the cognitive domain than the other groups because of (a) low expectations; (b) inadequately trained teachers; (c) full inclusion practices that ignored academic instructional needs; or (d) a combination of those and other, unknown, confounding variables.

The large effect size differences between those with MID and school-age students in the other two high-incidence disability categories can also be explained by changes over time in the population identified as having MID. More than 20 years ago, Polloway and Smith (1983) reported that individuals with MID were now "patently more disabled" than those so labeled previously. This means that students with MID are now (i.e., post-Polloway & Smith's comments) in need of greater functional and life skills, personal care skills, communication ability, social skills, and transition-oriented emphases in their education. At least compared with students with LD or pupils with E/BD, the present meta-analysis results verify the phenomenon that Polloway and Smith first reported in the early 1980s. Perhaps an array of placement options (i.e., the full continuum of placement alternatives) may be necessary to meet those needs.

In an interesting discussion related to labeling in special education, MacMillan, Siperstein, and Gresham (1996) suggested changing the term mild mental retardation to generalized learning disability when referring to students with MID. The authors recommended this change because they believed that students with MID were more similar to those with LD than they were to other students with more severe levels of intellectual disability. Except in the behavior domain, the data reported here do not support the contention that students with LD and MID are similar. Individuals with MID have starkly different cognitive characteristics from those with specific LD, and the studies synthesized here confirm this disparity.

The present results confirm that students with LD, when compared with those in other categories of high-incidence disability, demonstrate the highest levels of functioning in the three domains of our analysis. Students with LD were found to be more successful than those with MID and E/BD in two thirds of the group comparisons. Although statistically significant differences exist between those with LD and the other two groups in IQ, academic achievement, and behavior, one must question the practical significance of these findings. That is, conquering academic and behavior problems must continue to be paramount in educational settings for students with LD. The learning and behavior needs of students with LD will be appropriately served only when they receive explicit, intensive instruction, using appropriately designed curricula, for a sufficient duration.

The present academic achievement findings of students with E/BD relative to those with LD and pupils with MID are also generally in accord with other research (see Trout, Nordness, Pierce, & Epstein, 2003). Trout et al. also noted that the amount of research on academic achievement of students with E/BD has (a) declined in recent years, (b) predominantly examined students in restrictive educational environments, and (c) rarely measured potentially important modifier variables such as socioeconomic status. It is therefore essential to increase research attention concerning the educational achievement of Students with E/BD.

Present findings on behavior and emotional problems were based on rating scales completed by teachers or parents. Our results show that the average problem behavior score of students with E/BD was worse than a significant number (approximately 76%) of students with LD and youth with MID. These findings, however, could be specific to the measurement system used in the studies included in our analysis. It follows that research is needed to investigate emotional and behavioral problems of students with high-incidence disabilities as measured by a variety of methods including technically adequate self-reports, objective personality tests, sociometric methods, target behavior recording, as well as rating scales completed by adults. Another consideration about rating scales is that they have a restricted measurement range and may be prone to ceiling effects. If so, other measurement methods (e.g., target behavior recording in terms of frequency or duration) might reveal even greater differences in behavior problems between students with E/BD and those in the two other categories.

A limitation of the present analysis that should lead to caution when interpreting these findings concerns the "file drawer problem" found in meta-analysis reports or any other review of research literature. This dilemma concerns the extreme view that journals are filled with the 5% of studies that show significant differences in the results, whereas the file drawers of researchers are filled with the other 95% of studies that show nonsignificant findings. Although a statistical technique to overcome this weakness does not exist, calculations are available to indicate the minimum number of studies (or "fail-safe N") necessary for a meta-analysis to avoid the file drawer problem (see Orwin, 1983). Extrapolating Orwin's calculations to the present data, one would need 93 studies with a [[bar.ES].sub.w] of zero to reduce our [[bar.ES].sub.w] to .5 when contrasting students with LD versus those with MID on IQ, and 37 studies when comparing pupils with MID to participants with E/BD on IQ with similar statistical parameters. As can be seen in the present study, for the largest obtained [[bar.ES].sub.w] estimates, we had fewer than the necessary number of investigations, so the file drawer problem is plausible. Moreover, we may have inflated the file drawer problem by excluding unpublished research (e.g., dissertations) examining the same three factors across the three groups. Our analysis also did not categorize the published research into "high" or "low" quality because of the difficulty in professional agreement over what is quality empiricism (see McGuire et al., 1985; Wortman, 1994). The study inclusion criteria did not allow for studies conducted outside the United States. Similar studies do exist (in Israel; see Margalit, 1989, 1995; Margalit & Roth, 1989), and it is likely that including these additional international studies would have affected the ESs shown here.

Another weakness that could have contributed to the present results concerns the definitions and eligibility criteria used across states to identify students with LD (see Kauffman, Hallahan, & Lloyd, 1998), MID (see Denning, Chamberlain, & Polloway, 2000), and E/BD (see Forness & Knitzer, 1992). It is safe to say that great disparity exists in definitions and eligibility across states. The student groups who comprised the research samples in each category of high-incidence disability were far from homogeneous across studies, and that could have led to discrepant results in the 58 studies. These difficulties in design and execution of our analysis, therefore, suggest prudence in considering the power of the present findings; meaningless statistics are still a consequence of a meta-analysis of poorly executed studies or restrictions in study selection.

To some extent the uncertainty over the justification for separate categories of students with learning, intellectual, emotional, and behavioral problems involves the following question: Are students with LD, MID, and E/BD alike or different? Nearly everyone recognizes that students with E/BD vary considerably among themselves on school-related characteristics, as do students with LD and youth with MID. Therefore the question can be restated: Are the average differences across students with LD, MID, and E/BD greater than the differences among students within any one of the three categories? The present meta-analysis results clearly point to a "yes" answer with respect to intelligence, academic achievement, and behavior problems.

In a classroom serving all three types of students with high-incidence disabilities, we believe our results translate into differentiated educational content focus from the teacher. This does not imply, however, that differentiated teaching behaviors are required across the three categories because we believe that "effective instruction is effective instruction" no matter who receives it. Our analysis is insufficient without additionally knowing the differential effects of instruction across the three groups. Perhaps the most significant validation for multiple categories of high-incidence disabilities would be that knowing a student's category tells the practitioner how to intervene with that student. Separate special education categories "A," "B," and "C" would be useful if educators knew that students in each category generally respond differentially to a particular intervention. One way to know this is if investigators could point to research showing that any student in category A is likely to benefit from a particular intervention, but any student in categories B or C will not. Such inquiry in special education appears to be sparse at present. Our meta-analysis justifies the use of three different groups of high-incidence disabilities based on research, but interventions and services must continue to address documented, individual student need regardless of category.


Formulas (all originate with Lipsey & Wilson, 2001) Used in the Meta-Analysis

Cohen's d was calculated as: d = [M.sub.gr1] - [M.sub.gr2]/[SD.sub.pooled] where [M.sub.gr1] = mean score of either the group with LD or those with MID, on a dependent variable. In order to standardize all the obtained ESs in the two group comparisons involving those with LD, participants identified as LD were always considered Group 1. In comparisons involving only those with MID and E/BD, the group with MID was classified as Group 1. In other words, those identified as E/BD were not considered Group 1 in any ES statistic calculation. [M.sub.gr2] = mean score of either the group with MID or those with E/BD, on a dependent variable. The group with MID was always Group 2 when compared with youth with LD. The pooled standard deviation was calculated as:

[SD.sub.pooled] = [square root of ([n.sub.gr1] - 1)[SD.sup.2.sub.gr1] + (([n.sub.gr2] - 1)[SD.sup.2.sub.gr2]]/ ([n.sub.gr1 - 1]) + ([n.sub.gr2] -1])

where [n.sub.gr1] = sample size of Group 1; [n.sub.gr2] = sample size for Group 2; [SD.sub.gr1] = standard deviation for a dependent measure score for Group 1; and [SD.sub.gr2] = standard deviation for a dependent measure score for Group 2.

The standard error (SE) of the ES for each study was calculated as:

SE [SD.sub.pooled] [square roof of 1/[n.sub.gr1] + 1/[n.sub.gr2]]

The inverse variance weight (IVW) of the ES for each study was calculated as:

IVW = 1/[SE.sup.2] = [n.sub.gr1] [n.sub.gr2]/[SD.sub.pooled] ([n.sub.gr1 + [n.sub.gr2]

The inverse variance weight times effect size (IVW x ES) datum for each study was calculated simply by multiplying the study INV by the within-study ES.

The weighted mean effect size ([ES.sub.w]) for each study domain and two group comparison was calculated as: [[bar.ES.]sub.w] = [summation](IVWxES)/[summation] IVW]. In this equation, each study's IVW x

ES datum in each domain and two group comparison was summed and then divided by the sum of all the study IVWs for a domain and two group comparison. In other words, each domain and two group comparison had only one [[bar.ES].sub.w]. The [[bar.ES].sub.w] (versus the unweighted [bar.x] ES) for a domain and two group comparison considers all the contributing effect sizes with respect to sample size. Studies with larger sample sizes (versus smaller ns) provide a more robust estimate of the population parameters and are therefore assigned greater weight.

The domain and two group comparison Q statistic was calculated as: Q = [summation] IVW [(ES - [[bar.ES].sub.w]).sup.2]. In this equation, the single [[bar.ES].sub.w] for each two group comparison in each domain was subtracted from each study's single (within study) ES, and this difference was squared. Then, the squared difference was multiplied by the IVW of each study, then summed across studies, and called the Q statistic.

The standard error of the [[bar.[ES.sub.w] was calculated as: SE [[bar.ES].sub.w] = [square roof of 1/[summation] IVW]

The 95% confidence intervals around the [[bar.ES].sub.w] were calculated as: [[bar.ES].sub.w] (lower limit) = [[bar.ES].sub.w] - z(1-.05) (SE [[bar.ES].sub.w]), and [[bar.ES].sub.w] (upper limit) = [[bar.[ES.sub.w] + z(1-.05) (SE [[bar.ES].sub.w]). "To construct the confidence intervals, you multiply the SE [[bar.ES].sub.w] by a critical z-value representing the desired confidence level (in our case 1.96 for [alpha] = .05), and add the product to the [bar.[ES.sub.w] for the upper limit, and subtract the product from the [[bar.ES].sub.w] for the lower limit" (Lipsey & Wilson, 2001, p. 114). If the confidence interval does not include zero, the [[bar.ES].sub.w] (and the difference between groups) is statistically significant at the chosen alpha level.

The t-test standardized mean difference ES was calculated as: d = t [square root of [n.sub.1 + [n.sub.2]/[n.sub.1][n.sub.2]]

The F-ratio standardized mean difference ES was calculated as: d = [square root of F([n.sub.1 + [n.sub.2]/[n.sub.1] [n.sub.2]]

Manuscript received April 2004; manuscript accepted November 2004.


References marked with an asterisk indicate studies included in the meta-analysis.

Algozzine, B., & Ysseldyke, J. E. (1983). Learning disabilities as a subset of school failure: The oversophistication of a concept. Exceptional Children, 50, 242-246.

* Anderson, J. A., Kutash, K., & Duchnowski, A. J. (2001). A comparison of the academic progress of students with EBD and students with LD. Journal of Emotional and Behavioral Disorders, 9, 106-115.

Bateman, B. D. (1994). Who, how, and where: Special education's issues in perpetuity. The Journal of Special Education, 27, 509-520.

* Boyle, J. R., & Weishaar, M. (2001). The effects of strategic notetaking on the recall and comprehension of lecture information for high school students with learning disabilities. Learning Disabilities Research, 16, 133-141.

* Bramlett, R. K., Smith, B. L., & Edmonds, J. (1994). A comparison of nonreferred, learning disabled, and mildly mentally retarded students utilizing the social skills rating system. Psychology in the Schools, 31, 13-19.

* Bussing, R., Zima, B. T., Belin, T. R., & Forness, S. R. (1998). Children who qualify for LD and SED programs: Do they differ in level of ADHD symptoms and comorbid psychiatric conditions. Behavioral Disorders, 23, 85-97.

* Carroll, J. L., Friedrich, D., & Hund, J. (1984). Academic self-concept and teachers' perceptions of normal, mentally retarded, and learning disabled elementary students. Psychology in the Schools, 21, 343-348.

* Cohen, M., & Hynd, G. (1986). The Conners Teacher Rating Scale: A different factor structure with special education children. Psychology in the Schools, 23(1), 13-23.

Consortium for Citizens with Disabilities. (2002). Comments on the recommendations of the President's Commission on Excellence in Special Education. Washington, DC: Author

Cooper, H., & Hedges, L. V. (Eds.). (1994). The handbook of research synthesis. New York: Russell Sage Foundation.

* Costenbader, V. K., & Keller, H. R. (1990). Behavioral ratings of emotionally handicapped, learning disabled, and nonreferred children: Scale and source consistency. Journal of Psychoeducational Assessment, 8, 485-496.

* Cullinan, D., & Epstein, M. H. (1985). Adjustment problems of mildly handicapped and nonhandicapped students. Remedial and Special Education, 6(2), 5-11.

* Cullinan, D., Schultz, R. M., Epstein, M. H., & Luebke, J. F. (1984). Behavior problems of handicapped adolescent female students. Journal of Youth and Adolescence, 13, 57-64.

* Cunningham, S. (1988). Performance comparisons of learning disabled, nonhandicapped, and educable mentally retarded students on selected industrial arts tasks. Journal of Industrial Teacher Education, 25(3), 49-55.

Davies, H. T. O., & Crombie, L. K. (2003, May). What is meta-analysis? What is series, 1(8). Retrieved March 1, 2004, from ebmfiles/WhatisMetaAn.pdf

Denning, C. B., Chamberlain, J. A., & Polloway, E. A. (2000). An evaluation of state guidelines for mental retardation: Focus on definition and classification practices. Education and Training in Mental Retardation and Developmental Disabilities, 35, 119-134.

* Ellen, A. S. (1989). Discriminant validity of teacher ratings for normal, learning-disabled, and emotionally handicapped boys. Journal of School Psychology, 27, 15-25.

* Epstein, M. H., & Cullinan, D. (1983). Academic performance of behaviorally disordered and learning-disabled pupils. The Journal of Special Education, 17, 303-307.

* Epstein, M. H., Cullinan, D., & Gadow, K. D. (1986). Teacher ratings of hyperactivity in learning-disabled, emotionally disturbed, and mentally retarded children. The Journal of Special Education, 20, 219-229.

* Epstein, M. H., Polloway, E. A., Foley, R. M., & Patton, J. R. (1990). Comparisons of performance on academic probes by students with mild retardation, learning disabilities, and behavior disorders. Special Services in the Schools, 6, 121-134.

* Epstein, M. H., Polloway, E. A., Foley, R. M., & Patton, J. R. (1993). Homework: A comparison of teachers' and parents' perceptions of the problems experienced by students identified as having behavioral disorders, learning disabilities, or no disabilities. Remedial and Special Education, 14(5), 40-50.

Forness, S. R., & Knitzer, J. (1992). A new proposed definition and terminology to replace "serious emotional disturbance" in Individuals with Disabilities Education Act. School Psychology Review, 21, 12-20.

* Frank, A. R., Sitlington, P. L., & Carson, R. R. (1995). Young adults with behavioral disorders: A comparison with peers with mild disabilities. Journal of Emotional and Behavioral Disorders, 3, 156-164.

* Fuchs, D., Fuchs, L. S., & Power, M. H. (1987). Effects of examiner familiarity on LD and MR students' language performance. Remedial and Special Education, 8(4), 47-52.

* Fulk, B. M., Brigham, F. J., & Lohman, D. A. (1998). Motivation and self-regulation. Remedial and Special Education, 19, 300-309.

* Fuller, G. B., & Goh, D. S. (1981). Intelligence, achievement, and visual motor performance among learning disabled and emotionally impaired children. Psychology in the Schools, 18, 261-268.

* Gajar, A. H. (1979). EMR, LD, ED: Similarities and differences. Exceptional Children, 45, 470-472.

* Gajar, A. H. (1980). Characteristics across exceptional categories: EMR, LD, and ED. The Journal of Special Education, 14, 165-173

* Greene, A. C., Sapp, G. L., & Chissom, B. (1990). Validation of the Stanford-Binet Intelligence Scale: Fourth Edition with exceptional black male students. Psychology in the Schools, 27 (1), 35-41.

* Gresham, F. M., Elliott, S. N., & Black, F. L. (1987). Factor structure replication and bias investigation of the teacher rating of social skills. Journal of School Psychology, 25, 81-92.

* Gresham, F. M, MacMillan, D. L., & Bocian, K. M. (1996). Learning disabilities, low achievement, and mild mental retardation: More alike than different? Journal of Learning Disabilities, 29, 570-581. Hallahan, D. P., & Kauffman, J. M. (1977). Labels, categories, behaviors: ED, LD, and EMR reconsidered. The Journal of Special Education, 11, 139-149.

* Harris, J. C., King, S. L., Reifler, J. P., & Rosenberg, L. A. (1984). Emotional and learning disorders in 6-12 year-old boys attending special schools. Journal of the American Academy of Child Psychiatry, 23, 431-437.

Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. New York: Academic Press.

* Janke, R. W., & Lee, K. (1991). Social skills ratings of exceptional students. Journal of Psychoeducational Assessment, 9, 54-66.

Kauffman, J. K., Hallahan, D. P., & Lloyd, J. W. (1998). Politics, science, and the future of learning disabilities. Learning Disability Quarterly, 21, 276-280.

Kauffman, J. K., & Landrum, T. (2006). A brief history of special education for students with emotional and behavioral disorders. Austin, TX: PRO-ED.

* Kortering, L. J., & Braziel, P. M. (1999). Staying in school: The perspective of ninth-grade students. Remedial and Special Education, 20, 106-113.

* Lidz, C., & Ballester, L. E. (1986). Diagnostic implications of McCarthy Scale General Cognitive Index/Binet IQ discrepancies for low-socioeconomic-status preschool children. Journal of School Psychology, 24, 371-385.

Lipsey, M. W., & Wilson, D. B. (2001). Practical meta-analysis. Thousand Oaks, CA: Sage.

* Longman, R. S., Inglis, J., & Lawson, J. S. (1991). WISC-R patterns of cognitive abilities in behavior disordered and learning disabled children. Psychological Assessment, 3, 239-246.

* Lopez, M. F., Forness, S. R., MacMillan, D. L., Bocian, K. M., & Gresham, F. M. (1996). Children with attention deficit hyperactivity disorder and emotional or behavioral disorders in primary grades: Inappropriate placement in the learning disability category. Education and Treatment of Children, 19, 286-299.

* Luebke, J., Epstein, M. H., & Cullinan, D. (1989). Comparison of teacher-rated achievement levels of behaviorally disordered, learning disabled, and nonhandicapped adolescents. Behavioral Disorders, 15, 1-8.

Macmillan, D. L., Siperstein, G. N., & Gresham, F. M. (1996). A challenge to the viability of mild mental retardation as a diagnostic category. Exceptional Children, 62, 356-371.

Margalit, M. (1989). Academic competence and social adjustment of boys with learning disabilities and boys with behavior disorders. Journal of Learning Disabilities, 22, 41-50.

Margalit, M. (1995). Social skills learning for students with learning disabilities and students with behavior disorders. EducationaI Psychology, 15, 445-456.

Margalit, M., & Roth, Y. B. (1989). Strategic keyboard training and spelling improvement among children with learning disabilities and mental retardation. Educational Psychology, 9, 321-329.

Marston, D. (1987). Does categorical teacher certification benefit the mildly handicapped child? Exceptional Children, 53, 423-431.

* McConaughy, S. H., & Achenbach, T. M. (1996). Contributions of a child interview to multimethod assessment of children with EBD and LD. School Psychology Review, 25, 24-39.

McGuire, J., Bates, G. W., Dretzke, B. J., McGivern, E., Rembold, K. L., Seabold, D. R., et al. (1985). Methodological quality as a component of meta-analysis. Educational Psychologist, 20, 1-5.

* Merrell, K. W., Merz, J. M., Johnson, E. R., & Ring, E. N. (1992). Social competence of students with mild handicaps and low achievement: A comparative study. School Psychology Review, 21(1), 125-137.

* Merrell, K. W., Sanders, D. E., & Popinga, M. R. (1993). Teacher ratings of student social behavior as a predictor of special education status: Discriminant validity of the school social scales. Journal of Psychoeducational Assessment, 11, 220-231.

* Miniutti, A. A. (1991). Language deficiencies in inner-city children with learning and behavioral problems. Language, Speech, and Hearing Services in Schools, 22, 31-38.

* Naglieri, J. A. (1985). Use of the WISC-R and K-ABC with learning disabled, borderline mentally retarded, and normal children. Psychology in the Schools, 22 (2), 133-141.

* Obrzut, A., Obrzut, J. E., & Shaw, D. (1984). Construct validity of the Kauffman Assessment Battery for children with learning disabled and mentally retarded. Psychology in the Schools, 21, 417-424.

Orwin, R. G. (1983). A fail-safe N for effect size in meta-analysis. Journal of Educational Statistics, 8, 157-159.

Polloway, E. A., & Smith, J. D. (1983). Changes in mild mental retardation: Population, programs, and perspectives. Exceptional Children, 50, 149-159. President's Commission on Excellence in Special Education. (2002). A new era: Revitalizing special education for children and their families. Washington, DC: U.S. Department of Education.

* Reid, R., Epstein, M. H., Pastor, D. A., & Ryser, G. R. (2000). Strengths-based assessment differences across students with LD and EBD. Remedial and Special Education, 21, 346-355.

Reschly, D. J., & Till, W. D. (1999). Reform trends and system design alternatives. In D. J. Reschly, W. D. Tilly, & J. P. Grimes (Eds.), Special education in transition: Functional assessment and noncategorical programming (pp. 19-48). Longmont, CO: Sopris West.

Reschly, D. J., Tilly, W. D., & Grimes, J. (Eds.). (1999). Special education in transition: Functional assess ment and noncategorical programming. Longmont, CO: Sopris West.

Reynolds, M. C., & Birch, J. W. (1977). Teaching exceptional children in all America's schools. Reston, VA: Council for Exceptional Children.

* Richmond, B. O., & Blagg, D. E. (1985). Adaptive behavior, social adjustment, and academic achievement of regular and special education children. The Exceptional Child, 32, 93-98.

Rosenthal, R. (1991). Meta-analytic procedures for social research. Beverly Hills, CA: Sage.

Sabomie, E. J., Evans, R., & Cullinan, D. (in press). Comparing characteristics of high-incidence disability groups: A descriptive review. Remedial and Special Education.

* Sabornie, E. J., Kauffman, J. M., & Cullinan, D. (1990). Extended sociometric status of adolescents with mild handicaps: A cross-categorical perspective. Exceptionality, 1, 197-209.

* Sabornie, E. J., Kauffman, J. M., Ellis, E. S., & Marshall, K. J. (1987-1988). Bi-directional and cross-categorical social status of learning disabled, behaviorally disordered, and nonhandicapped adolescents. The Journal of Special Education, 21, 39-56.

* Sabornie, E. J., Marshall, K. J., & Ellis, E. S. (1988). Behavior disordered, learning disabled, and nonhandicapped students' social status in mainstream classes. Monograph in Behavior Disorders: Severe Behavior Disorders of Children and Youth, 11, 32-45.

SAS Institute. (2001). JMP (Version 4.0) [Computer software and manual]. Cary, NC: Author.

* Schachar, R., Logan, G., Wachsmuth, R, & Chajczyk, D. (1988). Attaining and maintaining preparation: A comparison of attention in hyperactive, normal, and disturbed control children. Journal of Abnormal Child Psychology, 16, 361-378.

* Schloss, P. J., Epstein, M. H., & Cullinan, D. (1988). Depression characteristics among mildly handicapped students. Journal of the Multihandicapped Person, 1, 293-302.

* Scott, M. S., Perou, R., Greenfield, D., Partridge, M. F., & Swanson, L. J. (1993). A comparison of normally achieving, mildly retarded and learning disabled students on a perceptually based oddity task. Journal of Developmental and Physical Disabilities, 5, 129-150.

* Scruggs, T. E., & Mastropieri, M. A. (1986). Academic characteristics of behaviorally disordered and learning disabled students. Behavioral Disorders, 11, 184-190.

Shadish, W. R., Robinson, L., & Lu, C. (1999). ES: A computer program and manual for effect size calculation (Version 1.0) [Computer software and manual]. St. Patti, MN: Assessment Systems Corporation.

* Sitlington, P. L., Frank, A. R., & Carson, R. (1992). Adult adjustment among high school graduates with mild disabilities. Exceptional Children, 59, 221-233.

* Slate, J. R., & Saarnio, D. A. (1995). Differences between WISC-III and WISC-R IQs: A preliminary investigation. Journal of Psychoeducational Assessment, 13, 340-346.

* Sprafkin, J., & Gadow, K. (1987). An observational study of emotionally disturbed and learning-disabled children in school settings. Journal of Abnormal Child Psychology, 15, 393-408.

* Sprafldn, J., Kelly, E., & Gadow, K. D. (1987). Reality perceptions of television: A comparison of emotionally disturbed, learning disabled, and nonhandicapped children. Journal of Developmental and Behavioral Pediatrics, 8(3), 149-153.

* Stavrou, E. (1990). The long-term stability of WISCR scores in mildly retarded and learning-disabled children. Psychology in the Schools, 27, 101-110.

Tilly, W. D., Reschly, D. J., & Grimes, J. (1999). Disability determination in problem-solving systems: Conceptual foundations and critical components. In D. J. Reschly, W. D. Tilly, & J. P. Grimes (Eds.), Special education in transition: Functional assessment and noncategorical programming (pp. 285-301). Longmont, CO: Sopris West.

Trout, A. L., Nordness, P. D., Pierce, C. D., & Epstein, M. H. (2003). Research on the academic status of children with emotional and behavioral disorders: A review of the literature from 1961 to 2000. Journal of Emotional and Behavioral Disorders, 11, 198-210.

* Vance, H. B., Fuller, G. B., & Ellis, R. (1983). Discriminant function analysis of LD/BD children. Journal of Clinical Psychology, 39, 749-753.

* Whorton, J. E. (1985). Test-retest Wechsler intelligence scale for children Revised scores for 310 educable mentally retarded and specific learning disabled students. Psychological Reports, 56, 857-858.

* Wilson, L., Cone, T., Bradley, C., & Reese, J. (1986). The characteristics of learning disabled and other handicapped students referred for evaluation in the state of Iowa. Journal of Learning Disabilities, 19, 553-557.

* Wolman, C. (1991). Sensitivity to causal cohesion in stories by children with mild mental retardation, children with learning disabilities, and children without disabilities. The Journal of Special Education, 25, 135-154.

* Wolman, C., van den Broek, P., & Lorch, R. F. (1997). Effects of causal structure on immediate and delayed story recall by children with mild mental retardation, children with learning disabilities, and children without disabilities. The Journal of Special Education, 30, 439-455.

* Wood, T. A., Buckhalt, J. A., & Tomlin, J. G. (1988). A comparison of listening and reading performance with children in three educational placements. Journal of Learning Disabilities, 21, 493-496.

Wortman, P. M. (1994). Judging research quality. In H. Cooper & L. V. Hedges (Eds.), The handbook of research synthesis (pp. 97-I 10). New York: Russell Sage Foundation.

* Wynne, M. E., & Brown, R. T. (1984). Assessment of high incidence learning disorders: Isolating measures with high discriminant ability. School Psychology Review, 13, 231-237.

Ysseldyke, J., & Marston, D. (1999). Origins of categorical special education services in schools and a rationale for changing them. In D. J. Reschly, W. D. Tilly, & J. P. Grimes (Eds.), Special education in transition: Functional assessment and noncategorical programming (pp. 1-18). Longmont, CO: Sopris West.





North Carolina State University

EDWARD J. SABORNIE (CEC Chapter #639), Professor; DOUGLAS CULLINAN (CEC Chapter #617), Professor; SUSAN S. OSBORNE (CEC Chapter #617), Associate Professor; and

LYNHE B. BROCK (CEC Chapter #639), Doctoral Student, Department of Curriculum and Instruction, North Carolina State University, Raleigh.

Address correspondence to Edward J. Sabornie, North Carolina State University, College of Education, Graduate Program in Special Education, Department of Curriculum and Instruction, Poe Hall, Campus Box 7801, Raleigh, NC 27695-7801. (e-mail:

We thank David Wilson, George Mason University, for his expert assistance in the technical and statistical preparation of this meta-analysis.
Effect Sizes and Related Meta Analysis Data


Data                           LD            LD          MID

                              vs.            vs.         vs.

Categories                    MID            E/BD        E/BD

[[bar.ES].sub.w]               2.36 *       0.14        -2.17 *

unweighted [bar.X] ES          2.58         0.06        -2.15

95% CI for               (low) 1.88        -0.55        -2.96
  [[bar.ES].sub.w]      (high) 2.83         0.84        -1.38

n of studies                  25           26           11

Q statistic (df)               5.34 (24)    1.66 (25)     .48 (10)

                          Domain and Group Comparisons
                             Academic Achievement

Data                     LD            LD          MID

                         vs.           vs.          vs.

Categories               MID          E/BD         E/BD

[[bar.ES].sub.w]          .42 *         .22 *       -.70 *

unweighted [bar.X] ES    1.13        -0.32        -1.07

95% CI for               0.18         0.16        -1.12
  [[bar.ES].sub.w]       0.66         0.28        -0.28

n of studies            12           15            6

Q statistic (df)         1.17 (11)   24.91 (14)     .99 (5)


Data                     LD            LD          MID

                         vs.           vs.          vs.

Categories               MID          E/BD         E/BD

[[bar.ES].sub.w]        -0.01          .72 *        .70 *

unweighted [bar.X] ES    0.13         0.65         0.69

95% CI for              -0.26         0.50         0.37
  [[bar.ES].sub.w]       0.24         0.94         1.03

n of studies            12           19           10

Q statistic (df)         1.03 (11)    2.08 (18)    1.97 (9)

Note. LD = learning disabilities; MID = mild intellectual
disabilities; E/BD = emotional/behavioral disabilities; ES = effect
size; CI = confidence interval; [[bar.ES].sub.w] = weighted mean
effect size; [bar.X] = mean; df = degrees of freedom.

* p < .05
Gale Copyright:
Copyright 2005 Gale, Cengage Learning. All rights reserved.