Lack of proportional representation of culturally, linguistically, and economically diverse (CLED) students in gifted programs is a critical issue in education. The common conception of a high-ability student as an excellent classroom student with high grades and exceptional achievement on standardized tests ignores the latent potential in students and considers only mani-fest abilities (Briggs, Reis, & Sullivan, 2008). High ability and potential masked by socioeconomic and cultural factors can go undiscovered and underdeveloped (Briggs et al., 2008).
The lack of proportional representation of traditionally underrepresented groups in gifted programs within schools is likely the result of the identification methods that fail to accurately detect all students with high potential, especially students from diverse backgrounds coupled with inequities in opportunity. Thus, we conducted a meta-analysis to explore how the use of identification methods influence the proportional representation of Black, Hispanic, and Native American students as gifted.
Definitions of Giftedness and Identification Methods
The principles of education rest on the idea of giving chil-dren what they need to be successful and achieve whatever dream they might have (Tyler, 1949). In other words, educa-tion should address the diverse needs of the students. Therefore, educators are morally bound to provide gifted and talented students with appropriately challenging educational
opportunities to help them realize their potentials (Renzulli, & Reis, 1991).
Giftedness is defined by the National Association for Gifted Children (NAGC) (n.d.) as:
Gifted individuals are those who demonstrate outstanding levels of aptitude (defined as an exceptional ability to reason and learn) or competence (documented performance or achievement in top 10% or rarer) in one or more domains. Domains include any structured area of activity with its own symbol system (e.g., mathematics, music, language) and/or set of sensorimotor skills (e.g., painting, dance, sports). (Para 5)
This definition of giftedness extends to include a larger concept of talents from students from a wide variety of back-grounds and cultures in gifted education (Gentry, 2009). Under this definition, a child can be considered gifted if he or she shows talent in only one area. This definition is far more inclusive than more traditional definitions, which rely on strict cutoff scores based on aptitude measures (Feldhusen & Jarwan, 2000; Van Tassel-Baska, 2005).
752107 GCQXXX10.1177/0016986217752107Gifted Child QuarterlyHodges et al.research-article2017
1Purdue University, West Lafayette, IN, USA
†These authors contributed equally to this work.
Corresponding Author:Jaret Hodges, Duke Talent Identification Program, Duke University, 300 Fuller Street, Durham, NC 27701. USA.Email: [email protected]
A Meta-Analysis of Gifted and Talented Identification Practices
Jaret Hodges1†, Juliana Tay1†, Yukiko Maeda1, and Marcia Gentry 1
AbstractResearchers consider the underrepresentation of Black, Hispanic, and Native American students is largely due to the use of traditional methods of identification (i.e., IQ and standardized achievement tests). To address this concern, researchers created novel nontraditional identification methods (e.g., nonverbal tests, student portfolios, affective checklists). This meta-analysis of 54 studies, consisting of 85 effect sizes representing 191,287,563 students, provides evidence that nontraditional identification methods, while able to narrow the proportional identification gap between underrepresented (Black, Hispanic, and Native American) and represented (Asian and White American) populations, are still unable to address the issue of education inequity. An overall risk ratio of 0.34 was calculated for nontraditional methods of identification in comparison with a 0.27 risk ratio for traditional methods. While the nontraditional methods help identify more underrepresented students as gifted, the results of this meta-analysis show that better identification methods are needed to address inequities in identification.
Keywordsgifted, identification, equitability, testing, underrepresentation, Black, Hispanic, Native American
Note that we do not intend to discuss in-depth the nuances of different definitions and conceptions of giftedness, or how they affect the identification of gifted students in this study because an abundant body of literature has already addressed these issues (e.g., Borland, 2003; NAGC, 2010; Renzulli, 1978). However, it is important to point out the various defi-nitions adopted by states as these definitions have implica-tions for the identification process. For example, a state that has adopted a definition of giftedness focused on intellectual and cognitive abilities is more likely to have an identification process using standardized achievement tests and other forms of verbal assessments. Conversely, if a state has a defi-nition of giftedness that accounts for gifted potential, creativ-ity, and/or implications from socioeconomic differences, the identification process is likely to include nonverbal assess-ments and other potentially more inclusive methods of identification.
Variant Definitions of Giftedness Used by States. From the list of 50 state definitions of giftedness (NAGC, 2013), we found that 43 of the 50 states placed an emphasis on intellectual and academic abilities, whereas only half considered potential abilities as part of the definition of giftedness. As such, many schools still rely on use of traditional test scores as part of their identification processes. The In the State of States by the Council of State Directors of Programs for the Gifted (CSDPG) and NAGC (NAGC & CSDPG, 2015) pointed out that 17 states used IQ scores and 15 states used achievement scores as part of the selection criteria. However, it is impor-tant to note that 20 states also reported using a multiple crite-ria model for their evaluation process even though they did not provide any information on the specific criteria (NAGC & CSDPG, 2015).
The variation in the identification methods can be attrib-uted to the various definitions of giftedness, especially among adherents who support a definition based on a poten-tial of gifts versus those who support a definition based on a manifestation of gifts. For our study, traditional identifica-tion methods refer to the use of standardized tests of achieve-ment and ability, which include state-based achievement tests and IQ measures. In contrast, nontraditional methods refer to the use of assessments that have nonverbal and mul-tiple criteria components aligned with the inclusive defini-tion of giftedness of NAGC (n.d.). This categorization of the different methods of identification is what educators com-monly use in their identification practices (Krisel, 2012; Van Tassel-Baska, 2005).
The consideration of talent potential is not new to the field. Passow and Frasier (1994) suggested the incorporating potential as part of the identification process. They believed that doing so would help create a more inclusive model for gifted programming as compared with one heavily reliant on test scores that reflect manifested achievements. Feldhusen and Jarwan (2000) also urged a more comprehensive method of identification that looks beyond performance and includes
considerations related to “problems, weaknesses, and needs” (p. 279). Furthermore, Renzulli (1978) highlighted the need to consider on gifted behaviors and characteristics and not only relying on performance on cognitive ability tests.
Another area in which states differ in their definitions of giftedness is the inclusion of terms that reflect diversity of student populations, such as racial groups and socioeco-nomic status, in the definition. For example, some states, such as Florida, North Carolina, and Washington, explicitly incorporate language concerning different socioeconomic status in their definitions. Other states, such as Alaska, Kansas, and Nevada, focus their definitions mainly on chil-dren with intellectual and academic giftedness without using other descriptors (NAGC, 2013). Additionally, states, such as Colorado, Iowa, and Maryland, have broader definitions that include creativity and leadership skills. As such, identi-fication procedures among states may differ, ranging from more conservative philosophies of aptitude-based identifica-tion to liberal philosophies based on broader definitions of giftedness.
Controversies in Identification Processes Stemming From Defini-tions of Giftedness. Gifted education provides students with the gifted and talented services they require to attain optimal educational outcomes (Gentry, 2009; Renzulli & Reis, 1991). This education may take place in the form of acceleration, curriculum compacting, or enrichment programs in areas of interests (Gentry, 2009; Renzulli & Reis, 1991). Researchers have consistently demonstrated that, without proper gifted services, students will not achieve their academic potential and in many cases may underperform (Subotnik, Olszewski-Kubilius, & Worrell, 2011). Therefore, it is imperative for educators to accurately identify those students who need dif-ferentiated services. In other words, a transparent, research-based, and purposeful identification process is a critical first process in providing appropriate learning opportunities to gifted youth.
However, the processes for identifying gifted students have often been contentious (Giessman, Gambrell, & Stebbins, 2013; Lakin & Lohman, 2011; Lee & Olszewski-Kubilius, 2006; Lohman, 2005; Lohman, Korb, & Lakin, 2008; Naglieri & Ford, 2003, 2005; Renzulli & Reis, 2012). For example, how to identify, what to identify, and when to identify are some of the questions that plague the field (Callahan, 2005; Erwin & Worrell, 2012; Feldhusen & Jarwan, 2000; McClain & Pfeiffer, 2012; McKenzie, 1986; Passow & Frasier, 1994).
With the differences in defining giftedness, the identifica-tion of giftedness can be broadly classified into two catego-ries: one focusing on exhibited talent and the other on gifted potential (McKenzie, 1986; Pfeiffer & Blei, 2008). Researchers have focused on defining giftedness and, in turn, have devel-oped identification procedures to address issues of manifest talents and latent talents. Manifest talents are talents displayed and readily apparent to the observer, for example, high scores on aptitude and/or achievement tests or clearly displayed
Hodges et al. 149
precocious ability in a domain. In contrast, latent talents reflect unactualized potential that can be masked by environ-mental or social factors (e.g., the child who has the potential to be a musical prodigy but who has no access to a musical instrument).
Researchers (e.g., Pfeiffer & Blei, 2008; Sternberg & Davidson, 2005) have also debated whether identification should account only for the observed and measurable ability of a child or whether it should take into account a child’s nonmanifested potential. Traditionally, when exhibition of gifts has been considered, IQ scores have been used to define giftedness, with students scoring above a cutoff point being selected for the gifted programming (Lakin & Lohman, 2011; Peters & Gentry, 2012; Pfeiffer & Blei, 2008). The Wechsler Intelligence Scale for Children (WISC; Watkins, Greenawalt, & Marcell, 2002) is frequently used in the iden-tification process of giftedness in schools across the United States (McClain & Pfeiffer, 2012; McKenzie, 1986). In con-junction with cutoff scores from standardized achievement tests, these scores form the core of identification methods across the nation (NAGC & CSDPG, 2015). Using IQ tests as the sole instrument to select students for gifted programs has received much criticism in the field of gifted education, especially when doing so does not account for the recent changes in the definition of giftedness to include gifted potential and talent development (Krisel, 2012; Pfeiffer & Blei, 2008). As such, if schools are only using IQ scores to identify gifted students, Black, Hispanic, and Native American students who may not have the opportunities to develop their gifted potential are not likely to be identified and served.
Aside from the issue of using IQ as an assessment of gift-edness, there is also the issue of the validity of using IQ tests for identification. In particular, the use of IQ tests has been considered as one reason for the underrepresentation of gifted minorities (Pfeiffer, 2012). Pfeiffer and Blei (2008) cautioned against using IQ as a sole identification measure given its lack of context; whereas, D. Ford (1998) pointed out that many IQ tests are racially biased. Conversely, Erwin and Worrell (2012) noted that IQ tests might be biased, but that they measure exactly the constructs they are meant to measure and the question should be whether those constructs should be the sole traits that identify a child for gifted ser-vices. As IQ tests are verbal and quantitative, Black, Hispanic, and Native American students who do not have the chance to develop their abilities in these areas are not likely to be able to excel in these tests. Furthermore, with the high cutoff scores needed for students to be tested into gifted pro-grams, differences between Black, Hispanic, and Native American students and their peers only widen, making pro-portional representation difficult to achieve.
An alternative definition, one that considers abilities and talents in students may be underdeveloped but with adequate support can manifest, uses criteria such as aptitude, recom-mendations by teachers and peers, creativity, and nonverbal
assessments for identifying gifts (Lohman & Nicpon, 2012). Foremost among the nonverbal ability tests are the Naglieri Nonverbal Abilities Test (NNAT; Naglieri & Ford, 2003) and the Raven Standard Progressive Matrices (RAVEN; Raven, 2000), two of the most commonly administered tests by dis-tricts as a means for alternate identification. Raven (2000) highlighted the effect of culture and environment on intelli-gence when measured with the traditional standardized tests. Thus, he focused on two indicators of intelligence: “educa-tive ability” and “reproductive ability” (Raven, 2000, p. 2) as measures of intelligence in RAVEN. Naglieri and Ford (2003) also questioned the identification of gifted students through assessments of their academic abilities because they believed these assessments disadvantaged students who had limited verbal and quantitative skills. Rather they chose to focus on reducing cultural bias in their test items. Thus, NNAT and RAVEN focus on assessing students’ problem solving, reasoning, and observation skills and do not rely on language or cultural specific content knowledge (Lewis, DeCamp-Fritson, Ramage, McFarland, & Archwamety, 2007; Naglieri & Ford, 2003).
The NNAT has faced criticism from Lohman (2005) who questioned its effectiveness in identifying culturally and lin-guistically diverse students. Lohman claimed that the data presented by Naglieri and Ford (2003) did not corroborate their conclusions. In response Naglieri and Ford (2005) argued that more efforts could be spend on addressing the lack of representation rather than trying to undermine a tool that is used to help close the representation gap.
Some identification methods for giftedness combine ele-ments from traditional and nontraditional forms of assess-ment by including a nonverbal component in the testing. This is done in hope of reducing the language bias that may exist within traditional verbal and quantitative assessments. The Cognitive Abilities Test (CogAT; Lohman, 2011) is one such test. Currently in its seventh edition, the CogAT Form 7 uses verbal, quantitative, and nonverbal components to assess a student’s acquired abilities (Lohman, 2011). The inclusion of a nonverbal component was specifically designed to increase the identification of nontraditional gifted students.
Nevertheless, these standardized tests not only fail to silent the points of contention but their use also raises more questions about the reliability of the data and the validity of the inferences based on the data they yield (Lewis et al., 2007; Lohman et al., 2008). For example, when comparing students who were identified based on their achievement on standardized tests or parents’ nominations with their perfor-mance on the ACT (American College Testing), Lee and Olszewski-Kubilius (2006) found that Hispanic students identified through parents’ nominations did better on the ACT than students who were identified via standardized tests. Additionally, Giessman et al. (2013) compared stu-dents’ performances on CogAT, NNAT, and WISC and found that the tests yielded different rates of identification for gifted students. Furthermore, the rates of identification also
150 Gifted Child Quarterly 62(2)
differed among the various racial groups. For example, the difference in WISC scores between Black students and their White peers was 17.2 points.
In examining only nonverbal methods of identification, Lewis et al. (2007) conducted a study with 175 third to fifth graders and eighth graders in a midwestern city and com-pared the effectiveness of RAVEN, NNAT, and Iowa Test of Basic Skills (ITBS) in identifying Black, Hispanic, and Native American students. The researchers found that RAVEN identified a more ethnically diverse group of stu-dents than ITBS or NNAT. Interestingly, Lewis et al. (2007) found no difference in the proportion of Black, Hispanic, and Native American students identified by ITBS and NNAT, although ITBS as a traditional achievement test is expected to identify fewer underrepresented gifted students than the NNAT (i.e., nonverbal method). In addition, Lohman (2007) found that of the three components of CogAT, the nonverbal component is the least correlated to academic achievement and should not be used as a sole measurement for giftedness. Lohman et al. (2008) also raised questions about the validity of the nonverbal tests as an identification method in general. Using three different nonverbal tests, CogAT–Nonverbal (NV), NNAT, and RAVEN, they found much variability in the scores obtained by the elementary students, which would in turn influence identification rates. In particular, they noted that the NNAT and CogAT-NV both showed a statistically significant dif-ference in identification rates between English language learning (ELL) students and non-ELL students. Hence, even within the various nonverbal methods of identification, issues of validity are still being debated.
Another nontraditional method of identification involves using teachers to nominate children for services. Teacher nom-ination helps identify students who may not perform well on standardized achievement tests due to reasons such as lan-guage or cultural bias (Callahan & Miller, 2005; Renzulli, 2005). For example, the Having Opportunities Promotes Excellence (HOPE) Scale (Gentry, Peters, Pereira, McIntosh, & Fugate, 2015; Peters & Gentry, 2012) is an instrument that measures students’ academic and social characteristics of gift-edness as identified by their teachers. However, such nomina-tion processes by teachers have also been criticized for subjectivity and possible bias (McBee, 2006; Milner & Ford, 2007). Even if teachers are reliable identifiers of gifted stu-dents of similar cultural backgrounds, it is arguable whether they are able to reliably identify children from diverse back-grounds (McBee, 2006). Nevertheless, some researchers argue that teacher nominations should be included as one criterion or pathway in the identification process.
In the past decade, this issue of traditional versus nontra-ditional identification has often been debated among researchers (Giessman et al., 2013; Lakin & Lohman, 2011; Lohman et al., 2008; Naglieri & Ford, 2003). Although numerous researchers (Lakin & Lohman, 2011; Peters &
Gentry, 2012) have conducted research on the effectiveness of different test batteries in identifying students for gifted programs, no single way of identifying gifted students exists because the identification process may include the various issues highlighted above (Reis & McCoach, 2000).
Furthermore, there is the issue of equitable representation in gifted education, a position taken up recently by NAGC (2011). For equitable representation to exist, definitions of giftedness and identification methods must be congruent and concerned with equity. Should the definition focus only on measurable achievements or should it also take into consid-eration a student’s potential for growth? Similarly, is there a match between the identified students and the gifted services provided, and are the needs of the students being met (Peters, Matthews, McBee, & McCoach, 2013)? These are some questions that need to be addressed as the field moves toward achieving equitable representation.
Underrepresentation of Students in Gifted Programs
The lack of equitable representation in gifted programs has been an ongoing concern in the field of gifted education (Daniels, 1998; D. Y. Ford & Harris, 1994; Naglieri & Ford, 2003; Yoon & Gentry, 2009). As the population of the United States continues to diversify, the need for equitable represen-tation in gifted programs is paramount. Despite increasing rates of inclusion for minority students in gifted programs since the 1970s (Donovan & Cross, 2002), CLED students are still vastly underrepresented compared with their peers (Erwin & Worrell, 2012; D. Ford, 2014; Konstantopoulos, Modi, & Hedges, 2001; Yoon & Gentry, 2009).
For example, Konstantopoulos et al. (2001) reported the odds ratio (OR) for being identified as gifted for Black, Hispanic, and Native American students compared with Asian and White peers. They found that Black (OR = 0.37), Hispanic (OR = 0.45), and Native American Students (OR = 0.17) were all identified at lower rates compared with their White peers. Asian were identified at a higher rate than their White peers (OR = 2.17). Similarly, Yoon and Gentry (2009) found that Asian and White American students were over-represented proportionately in gifted programs throughout the United States. With an in-depth state-by-state analysis of the identification of Black, Hispanic, and Native American students for gifted services, Yoon and Gentry found that Black and Hispanic students were underrepresented propor-tionately in 47 states and Native American students were underrepresented in 43 states. The recent report by the Office for Civil Rights (OCR, 2015), which examined the overall representation of students within the U.S. public school sys-tem, supports Yoon and Gentry’s (2009) findings. Similar findings were also reported by D. Ford (2014), who used a Relative Difference in Composition Index to show that more than one-half million Black and Hispanic students remained
Hodges et al. 151
unidentified as gifted students. She suggested that if a certain proportion of a school were composed of a given minority, then that same proportion should be represented in its gifted programming.
Sullivan (2011) found that statistically significant predic-tors for admittance to gifted programs are not found within academic predictors; in fact, socioeconomic status and race can more accurately predict a child’s identification as a gifted student than academic predictors. Interestingly, in Utah, Warne, Anderson, and Johnson (2013) found that when con-trolling for academic achievement, there was proportional racial representation in that state in gifted programs; that is, Black, Hispanic, and Native American gifted students were not underrepresented.
Lack of proportional representation in gifted programs is a critical issue in education. In 2001, the lack of educational equity across the spectrum in U.S. public education led to the introduction of No Child Left Behind (NCLB; Bush, 2001). However, unequal representation of races in gifted education continues. Hopkins and Garrett (2010) pointed out that the overrepresentation of Asian and White American students while Black, Hispanic, and Native American students were underrepresented constitutes de facto segregation within public schools. For example, Texas has seen the proportion of Hispanic students enrolled in public schools increase without a similar increase in Hispanic students identified as gifted (Esquierdo & Arreguín-Anderson, 2012). Furthermore, many still consider a high-ability student as a model student with good grades, taking into account only abilities that can be seen and measured by standards and tests (Briggs et al., 2008). Consequently, students’ socioeconomic and cultural factors may mask high ability and potential resulting in the student’s ability remaining undiscovered and underdevel-oped by schools (Briggs et al., 2008).
The differing percentages of identified students from vari-ous racial groups in a district are likely to correlate with the decision making of the administrators (Raudenbush, Fotiu, & Cheong, 1998) because school administrators determine the identification methods used in districts. These decisions can affect not only how students are identified but also how financial resources are allocated to gifted education pro-grams (Kettler, Russel, & Puryear, 2015). Thus, identifica-tion of CLED students for gifted services promotes social equity within a school district, along with funding and staff allocations (Grantham, 2011; Kettler et al., 2015; Raudenbush et al., 1998). Grantham (2011) called for educators to be “upstanders” rather than just bystanders to the lack of pro-portional representation in gifted education. In the case of Black male students, he argued that allowing nearly 150,000 youth to remain unidentified for gifted services is socially unjust (Grantham, 2011).
Differential Identification Rates by Grade Level. Another con-cern in the field of gifted education is determining when chil-dren should be identified. Researchers have brought to
attention the importance of early identification of gifted chil-dren, especially concerning the children’s cognitive, motiva-tional, and social–emotional development (Heller & Hany, 2004; Perleth, Schatz, & Mönks, 2000; Shaklee, 1992). The National Association for the Education of Young Children (2009) also highlighted the immediate and delayed influence of experiences and environment on children’s development. As such, it is not surprising that 26 out of 33 states identified children for gifted services in kindergarten and the elemen-tary grades compared with 11 states that provided additional identification services during middle school and nine states during high school (NAGC & CSDPG, 2015).
However, early identification is not without its concerns. When examining the performance of elementary school stu-dents in a longitudinal study using ITBS and CogAT, Lohman and Korb (2006) found that students’ scores in reading and mathematics differed the most when they were tested in Grade 4 than when they were tested in Grades 6 and 9. This shows that students who are identified in Grade 4 based on their ITBS scores may not perform similarly in Grades 6 and 9.
Another consideration on early identification is about the form of gifted programming available to the identified stu-dents. Olszewski-Kubilius and Limburg-Weber (1999) high-lighted the differences in needs between gifted elementary and gifted middle school students. The requisite for advanced classes in preparation for high school and college, as well as real-life experiences, are more appropriate for the older stu-dents as they prepare for this stage in their development. In comparison, elementary gifted programs focus on accelerat-ing and enriching content (NAGC, 2015).
Differential Identification Rates by Geographic Location. Imple-menting a unified identification process and method across differing states is not an easy feat to achieve. The issue is further compounded by variations in state laws and regula-tions concerning identification methods. The southern region of the United States faced litigation over underrepresentation of Black, Hispanic, and Native American students in gifted programs and made changes to identification processes (Ste-phen, Dudley, & Karnes, 2012). In Lee v. Lee County Board of Education (2007), the U.S. District Court mandated that the state of Alabama change its gifted education policy due to underrepresentation of Black students. Following the Office of Civil Rights’ investigation of the underrepresenta-tion of Black students in South Carolina, that state imple-mented changes to its gifted programming, which led to increased equitable representation (Swanson, 2007).
In addition, demographic differences among the states exist. Part of the southwest region, Texas and Arizona have a higher population of Hispanic Americans than other regions. However, in recent years, the midwest portion of the United States has also seen a surge in the Hispanic populations (Brown & Lopez, 2013). In comparison, the southern states have a higher population of Black Americans. There is little research on the identification of Native American students
152 Gifted Child Quarterly 62(2)
and it is unclear how geographic location influences this group of students.
Purpose of the Study
Proportional representation of students in gifted program-ming is a goal of gifted education researchers and practitio-ners. Identification methods affect the proportion of students from different races being served in gifted programs. Differences among the nontraditional identification tests have also yielded a variety of opinions among researchers about their effectiveness in identifying underrepresented stu-dents. Thus, the goal of this study is to shed light on the underrepresentation of CLED gifted students, specifically Black, Hispanic, and Native American students in gifted education programs. More specifically, we used a meta- analytic technique to gather and synthesize accumulated evi-dence of the underidentification of Black, Hispanic, and Native American gifted students reported in literature com-pared with that of Asian and White students as well as to examine the influence of traditional and nontraditional iden-tification methods on the underidentification of these populations.
Given the differing viewpoints on the effectiveness of identification methods in identifying Black, Hispanic, and Native American gifted students and the number of studies that have been conducted in the field of gifted education, a meta-analysis was appropriate to synthesize the wide range of studies. This was done in order to effectively gauge the gap in identification rates through nontraditional testing of underrepresented students compared with their counterparts in gifted programs as well as to examine the differences in identification rates between identification processes. More specifically, this meta-analysis was conducted based on the hypothesis that the identification methods are partially accountable for the lack of proportional representation and thus act as a barrier to gifted services among Black, Hispanic, and Native American students. In our study, Black, Hispanic, and Native American students are grouped as underrepre-sented students in gifted programming; whereas, Asian and White American students are considered as represented stu-dents. Specific research questions are the following:
1. Research Question 1: How do the proportional identification rates for gifted program services of Asian and White American students versus rates of Black, Hispanic, and Native American students differ?
2. Research Question 2: How does the identification method (i.e., traditional vs. nontraditional), location, and/or grade level of students moderate the differ-ence in proportional representation between Black, Hispanic, and Native American students identified as gifted compared with the proportion of Asian and White American students identified as gifted?
Method
Study Identification
Target literature for this meta-analysis was limited to studies reported only in the United States because this study is framed around NCLB (Bush, 2001), a law adopted in the United States. In addition, we chose to focus on studies reported between 2002 and 2015, setting the dates to coincide with the educational shift that came with the induction of NCLB. Although the main focus of NCLB was on general students, the policy resulted in a reduced focus on gifted students and limited funding spent on their special needs (Gentry, 2006). Another influence of NCLB on gifted students was related to closing the achievement gaps among the various groups of students, including English language learners, those from dif-ferent racial groups, and those from different socioeconomic classes (Gallagher, 2004). Furthermore, we limited the studies to those conducted in K-12 education because identification instruments and processes are largely used on students when they are within the K-12 school systems (Heller & Hany, 2004; NAGC & CSDPG, 2015; Perleth et al., 2000).
The keywords we used in our search were gifted, identifi-cation, intelligence, I.Q., equity, testing, high-ability, tal-ented, representation, underrepresented, underrepresentation, minority, and measurements. Different combinations of key-words and all of the keywords were used in the search. Some of the keywords (e.g., gifted and talented) were also chosen as they are often used interchangeably within the field of gifted education.
The meta-analysis commenced first with a search for lit-erature using electronic databases including Google Scholar, ERIC, PsycINFO, Thesis and dissertation platforms such as ProQuest. These databases have a large inventory of pub-lished and unpublished educational and psychological research. We considered these databases as sufficient starting points to identify potential studies for our meta-analysis. Potential articles were selected for inclusion through data-base searches; this was followed by a review of references in these studies to identify additional studies for inclusion in the meta-analysis.
Inclusion and Exclusion Criteria. To be included in the meta-analysis, a study needed to meet the following four criteria:
(a) Studies conducted in the United States with K-12 students after 2002: Studies concerning populations outside the United States or those dealing with un-dergraduates/postgraduates and pre-K students were excluded.
(b) Studies that involved gifted students as their study sample: Studies that included in their sample both gifted and nongifted student were included. Given that the meta-analysis was concerned with the differ-ence in representation of Black, Hispanic, and Native
Hodges et al. 153
American gifted students in gifted programming and general populations, those studies that only contained information about the gifted students in their sample were excluded, unless they indicated the size and composition of the general population.
(c) Studies that reported the identification information by race: Studies that did not describe the racial make-up of their sample were excluded.
(d) Studies that reported sufficient and relevant statis-tics that computed the identification rate by race: Literature reviews and qualitative studies were ex-cluded. The proportion of Black, Hispanic, and Na-tive American students in the general population and gifted population (or data provided to derive them) was necessary for inclusion.
The initial electronic search yielded 7,746 studies potentially related to our study. After screening the abstracts using the inclusion and exclusion criteria above, the number of articles potentially included in the meta-analysis was reduced to 183. We also conducted a manual search for studies that reported in the four major journals in gifted education (Gifted Child Quarterly, Roeper Review, Journal of Advanced Academics, and Journal for the Education of the Gifted). This step was put in place to verify that we included any potential articles not found through the initial search with the electronic databases. Each article published in these journals between the years 2002 and 2015 (k = 1,526) was scanned and assessed individually for inclusion in the meta-analysis. Furthermore, we examined available education statistical databases (OCR, 2015; Texas Education Agency, 2015) and state reports on state education websites. The two additional search methods identified addi-tional 55 articles and reports for 238 potential articles identified after the three initial screenings. These articles were then exam-ined by reading the texts carefully to retrieve effect sizes using the provided data in the reports. With the additional checks and computation, 186 studies were eliminated due to missing infor-mation in reporting of population, study sample, or identifica-tion methods. Thus, the final set of the studies included in this meta-analysis consisted of 45 journal articles, 4 dissertations, and 5 state reports, for a total of 54 studies. Figure 1 shows the search process. From these 54 studies and reports, that together included 191,322,595 students, we obtained 92 effect sizes. After accounting for dependent effect sizes, 85 independent effect sizes were extracted. The additional seven effect sizes were retrieved from Texas’ Public Education Information Management System (PEIMS) and OCR. We used the effect sizes calculated from PEIMS and OCR only for descriptive pur-poses. Table 1 lists the characteristics of the studies included in the meta-analysis.
Coding Process and Coded Variables
Before coding articles, a coding sheet was drafted and agreed upon by the team of researchers consisting of two graduate
students, a researcher with expertise in gifted education, and a researcher with expertise in research methodology includ-ing meta-analysis. Following the development of a coding sheet, an initial pilot examination of 20 articles by the team led to revisions of the coding sheet for clarity. The revised coding sheet included authorship, geographic location of the sample, year of publication, grade level, what identification method the researchers of primary studies employed, demo-graphic data about the student sample, as well as quantitative information to be used for computing an effect size. In con-junction with the type of identification method, whether that method constituted a verbal or nonverbal method was also coded. These variables are described and summarized in Table 1 and are also considered as potential moderators in the subsequent inferential analyses.
Identification Methods. The identification methods reported in examined studies were first classified into two categories: nontraditional and traditional depending on the identification tests and procedures used. These methods were further sub-categorized by the specific test being used into three nontra-ditional methods (RAVEN, NNAT, and CogAT-NV) and two traditional methods (IQ tests and achievement tests). This coincides with the literature since these are the most com-monly used identification methods (Giessman et al., 2013; Lakin & Lohman, 2011; Lohman et al., 2008; NAGC & CSDPG, 2015; Naglieri & Ford, 2003).
Geographic Location. We coded the location where the data were collected using six regional categories (i.e., north, northwest, southwest, south, west, and midwest). However, no studies included samples from the northwest portion of the United States. In addition, only two were included from the West, an insufficient number for including it in a modera-tor analysis.
Grade Level. Students in grades kindergarten to Grade 5 were coded as elementary students (k = 33); studies that included students in kindergarten to Grade 8 were coded as elemen-tary to middle school students (k = 9); and Grades 7 to 12 as middle to high school students (k = 11). Studies that involved a student sample across all grade levels (K-12) were not included in moderator analysis for grade level. We coded in this manner to preserve information about elementary stu-dents and how identification and representation differed from upper grade levels. Testing for gifted services is primar-ily done during elementary years (Sternberg & Davidson, 2005), and this is evident given the number of studies involv-ing elementary students (k = 42).
Two members of the research team completed the coding. Digital copies of the articles were kept in a shared folder to provide research team members access to all articles. A com-mon coding problem encountered by the coding team was that majority of the identified studies did not provide a clear description of their participants, which necessitated further
154 Gifted Child Quarterly 62(2)
calculations to obtain the effect sizes. Studies needing calcula-tions were set aside after the first round of coding for further evaluation of coding validity. Once initial coding of all studies was completed, one member of the coding team analyzed the studies with missing data and completed all necessary compu-tations to conclude coding, while another member of the team worked on keying in the information into the data sheet. Both members referred to all identified articles to ensure that the information was recorded correctly, and any disagreement concerning the coding was discussed between coders. The third research team member confirmed the computation of all effect sizes. The coded data were cross-checked again by the
coders and another member of the research team before con-ducting inferential analyses.
Definition of Effect Size
Given that inclusion (either a student is identified as gifted or not) is a dichotomous outcome, a risk ratio (RR; Borenstein, Hedges, Higgins, & Rothstein, 2009) that compares the rates of inclusion in gifted program between underrepresented and rep-resented populations was deemed the most appropriate effect size for analysis and most meaningful for interpretation. First, the subgroups that had been designated by race were
Figure 1. Meta-analysis search process.
Hodges et al. 155
consolidated into two groups: traditionally underrepresented populations (i.e., Black, Hispanic, and Native American stu-dents), which served as our focal group, and traditionally over-represented populations (i.e., White and Asian students), which served as the reference group. The RR compares the occurrence of identification of gifts in the focal group compared with that in the reference group. More specifically, in our meta-analysis, the overall effect size is defined as:
Risk Ratio RRP
PURG
RG( ) =
where PURG
= the proportion of Black, Hispanic, or Native American students who were identified or in a gifted pro-gram, P
RG = the proportion of White or Asian students who
were identified or in a gifted program. We also computed the RR for comparing the proportion in a gifted program for a specific racial group (e.g., Black) with traditional group, using the above-described formula.
An RR greater than 1 indicated that the focal group (i.e., underrepresented group) is overrepresented in a gifted pro-gram relative to the reference group (i.e., overrepresented group); whereas, an RR less than 1 indicated underrepresen-tation of the focal group compared with the reference group. For example, an RR of 2 is interpreted as the probability of being identified for gifted is as twice as high for the focal group compared with that for the reference group.
Alternatively, an RR of 0.5 means that the probability of being identified for the focal group is a half of that in the reference group. An RR of 1 indicates that all demographics are equally represented in a gifted program in a primary study. Table 1 also reports the RRs obtained from studies used in the analysis.
Handling Multiple Effect Sizes. In some cases (k = 12), multi-ple effect sizes were extracted from a single journal article. For example, Giessman et al. (2013) examined how identifi-cation methods at different cutoff scores led to different lev-els of proportional identification. In this case, three different identification methods were examined at three cutoff scores leading to nine effect sizes. Biased parameter estimates asso-ciated with multiple effect sizes from the same study, when they are dependent of each other, have often been noted in meta-analysis literature (e.g., Wood, 2008). As remedies, several methods of handling dependent effect sizes were sug-gested including selecting one representative effect size from each uniquely identified sample (e.g., Card, 2012; Steenbergen-Hu & Olszewski-Kubilius, 2016), averaging the multiple effect sizes within the study to obtain a synthetic effect size (e.g., Sutton, Abrams, Jones, Sheldon, & Song, 2000), using shifting unit of analysis (Cooper, 2010), estimating depen-dency with three-level multilevel modeling (Van den Noort-gate, López-López, Marín-Martínez, & Sánchez-Meca, 2013, 2015), applying multivariate methods to analyze dependent effect sizes by modeling the covariance structure (e.g., Olkin & Gleser, 2009), using a robust variance
estimation method (e.g., Hedges, Tipton, & Johnson, 2010). Although the multivariate approach is likely to offer minimal estimation errors (Hedges et al., 2010), the method was not feasible in our study due to the relatively small k for applying the method and limitation of required additional statistical information, such as the covariance structure of residuals, Hedges et al. (2010) also claim that the method is rarely used in practice for the additional data requirements and the time-consuming process. Robust variance estimation method and three-level multilevel model approaches are attractive options for handling dependencies when clusters of interre-lated effect sizes are observed in primary studies However, Hedges et al.’s (2010) method also requires correlation esti-mates for any pairs of effect sizes within a primary study. Furthermore, the application of three-level multilevel model-ing will suffer low statistical power and biased estimation of variance components when the sample size is small. As our study has only seven studies that reported dependent effect sizes from 54 studies, and the application of either of these methods may not be feasible for the current study. Card (2012) noted that the method of selecting the most represen-tative effect size should be applied only when it is clear that one particular effect size should be included whereas others should not because the selection decision may be an addi-tional source of bias due to researchers’ subjectivity (Wood, 2008). Thus, we used the method of averaging the effect sizes and Cooper’s (2010) shifting units of analysis, depend-ing on the purpose of the analysis. Although we acknowl-edge our selections of handling the multiple effect sizes may also have limitations, Hedges et al. (2010) observes our methods have commonly utilized in practice.
Averaging the Effect Sizes. A single effect size for each identification method was extracted by averaging the effect sizes with their associated standard error (SE) as weights, assuming these represent, in some extent, the same popula-tion effect size within the study. Then, we included the syn-thetic effect sizes (the study-average effect sizes; Hedges et al., 2010) to compute the average effect size across stud-ies. For example, Lohman and Gambrell (2012) averaged two extracted effect sizes to handle the dependency.
Shifting Unit of Analysis. When calculating the average effect size by study characteristics (e.g., identification meth-ods), shifting the unit of analysis technique (Cooper, 2010) was employed to preserve k for each follow-up moderator analyses, while handing dependency. Giessman et al. (2013) applied shifting the unit of analysis method to compute the weighted average effect sizes by identification methods (i.e., verbal vs. nonverbal) to minimize the dependency. When conducting the moderator analysis with identification meth-ods, the average verbal effect size was then removed from the analysis where pertinent. This procedure was replicated for all cases in which verbal and nonverbal effect sizes were extracted from the same study.
156
Tab
le 1
. Su
mm
ary
of t
he S
tudy
Cha
ract
eris
tics
Incl
uded
in M
eta-
Ana
lysi
s.
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Bern
al (
2002
)35
2,74
72.
80/8
.40
0.40
/1.5
00.
20/0
.30
96.5
0/89
.90
0.31
(0.
02)
1.50
(0.
11)
0.32
(0.
02)
——
Sout
hwes
t36
3,27
02.
70/7
.70
0.50
/1.6
00.
20/0
.30
96.6
0/90
.40
0.32
(0.
02)
1.50
(0.
11)
0.33
(0.
02)
——
Sout
hwes
t37
3,25
12.
50/7
.00
0.60
/1.6
00.
20/0
.30
96.8
0/91
.00
0.32
(0.
02)
1.50
(0.
11)
0.33
(0.
02)
——
Sout
hwes
t38
3,46
02.
30/6
.30
0.50
/1.8
00.
20/0
.30
96.9
0/91
.50
0.33
(0.
02)
1.50
(0.
11)
0.34
(0.
02)
——
Sout
hwes
t48
,587
1.30
/2.9
20.
36/0
.78
0.15
/0.1
498
.16/
96.0
90.
44 (
0.09
)0.
93 (
0.47
)0.
46 (
0.09
)—
—So
uthw
est
Brac
ken
and
Brow
n (2
008)
795
45.2
0/80
.00
15.7
0/6.
209.
60/3
.10
30.5
0/10
.70
0.24
(0.
37)
0.30
(0.
73)
0.27
(0.
41)
Ver
bal (
Brac
ken
Basi
c C
once
pt
Scal
e–R
evis
ed)
+ C
linic
al
Ass
essm
ent
of
Beha
vior
–Tea
cher
ESo
uth
752
65.1
0/60
.10
11.8
0/20
.10
8.60
/7.0
013
.70/
12.1
00.
82 (
0.22
)0.
68 (
0.35
)0.
58 (
0.28
)N
onve
rbal
(N
NA
T)
+ C
linic
al
Ass
essm
ent
of
Beha
vior
–Tea
cher
ESo
uth
Brul
les,
Pet
ers,
and
Sa
unde
rs (
2012
)3,
716
18.1
3/31
.23
3.01
/7.2
27.
16/5
.78
69.8
9/54
.15
0.45
(0.
09)
0.80
(0.
19)
0.51
(0.
09)
Non
verb
al (
NN
AT
)E
+ M
Sout
hwes
t
Car
man
and
Tay
lor
(201
0)2,
072
61.1
0/67
.20
9.80
/15.
506.
40/3
.60
22.2
0/13
.70
0.52
(0.
13)
0.55
(0.
27)
0.56
(0.
15)
Non
verb
al (
NN
AT
)E
Sout
h
Cri
bbs
(200
9)14
,701
34.1
0/78
.30
2.40
/10.
0059
.30/
8.40
2.40
/3.3
00.
08 (
0.20
)0.
063
(0.2
3)1.
39 (
0.37
)V
erba
l (A
chie
vem
ent
test
)M
Mid
wes
t
Cro
ss, N
eum
eist
er,
and
Cas
sady
(2
007)
5,50
075
.60/
69.9
41.
30/7
.74
12.2
0/7.
006.
70/3
.21
0.49
(0.
15)
0.54
(0.
18)
0.46
(0.
25)
Ver
bal (
SAT
)H
Mid
wes
t
Cur
by, R
udas
ill,
Rim
m-K
aufm
an,
and
Kon
old
(200
8)
347
56.0
0/88
.30
—44
.00/
11.6
0—
0.17
(0.
41)
0.17
(0.
41)
—V
erba
l (W
oodc
ock
John
son
Ach
ieve
men
t T
est)
/Ver
bal
(Cog
nitiv
e A
bilit
ies
Tes
t)/V
erba
l (S
tanf
ord–
Bine
t In
telli
genc
e Sc
ale)
/Non
verb
al
(RA
VEN
)
E +
MSo
uth
Esqu
ierd
o an
d A
rreg
uín-
And
erso
n (2
012)
77,7
72,0
0068
.63/
67.6
94.
23/9
.40
12.5
3/9.
1514
.61/
12.7
90.
75 (
0.00
)0.
70 (
0.00
)0.
86 (
0.00
)—
—N
atio
nal
D. F
ord
(201
4)48
,300
,000
51.0
0/63
.00
4.00
/10.
0019
.00/
10.0
025
.00/
16.0
00.
45 (
0.00
)0.
47 (
0.00
)0.
57 (
0.00
)—
—N
atio
nal
(con
tinue
d)
157
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Gie
ssm
an e
t al
. (2
013)
5,83
362
.80/
79.5
05.
10/1
3.30
20.8
0/4.
404.
90/2
.70
0.22
(0.
12)
0.18
(0.
15)
0.54
(0.
19)
Non
verb
al (
NN
AT
)E
Mid
wes
t5,
833
62.8
0/75
.60
5.10
/18.
9020
.80/
3.60
4.90
/2.0
00.
17 (
0.18
)0.
14 (
0.22
)0.
40 (
0.30
)N
onve
rbal
(N
NA
T)
EM
idw
est
5,83
362
.80/
72.9
05.
10/2
3.70
20.8
0/2.
404.
90/2
.10
0.14
(0.
28)
0.09
(0.
38)
0.42
(0.
41)
Non
verb
al (
NN
AT
)E
Mid
wes
t5,
833
62.8
0/77
.90
5.10
/12.
6020
.80/
6.60
4.90
/2.8
00.
30 (
0.10
)0.
27 (
0.12
)0.
56 (
0.19
)N
onve
rbal
(C
ogA
T)
EM
idw
est
5,83
362
.80/
76.8
05.
10/1
5.40
20.8
0/5.
504.
90/2
.30
0.24
(0.
16)
0.22
(0.
18)
0.46
(0.
28)
Non
verb
al (
Cog
AT
)E
Mid
wes
t5,
833
62.8
0/78
.50
5.10
/17.
6020
.80/
2.80
4.90
/1.1
00.
12 (
0.30
)0.
11 (
0.36
)0.
22 (
0.57
)N
onve
rbal
(C
ogA
T)
EM
idw
est
5,83
362
.80/
86.0
05.
10/8
.90
20.8
0/2.
804.
90/1
.50
0.13
(0.
15)
0.11
(0.
18)
0.30
(0.
25)
Ver
bal (
Cog
AT
)E
Mid
wes
t5,
833
62.8
0/86
.70
5.10
/9.5
020
.80/
2.00
4.90
/1.7
00.
11 (
0.22
)0.
08 (
0.30
)0.
34 (
0.33
)V
erba
l (C
ogA
T)
EM
idw
est
5,83
362
.80/
86.1
05.
10/1
1.20
20.8
0/1.
704.
90/1
.00
0.08
(0.
36)
0.06
(0.
45)
0.20
(0.
59)
Ver
bal (
Cog
AT
)E
Mid
wes
tIll
inoi
s St
ate
Boar
d of
Edu
catio
n (2
003)
2,08
4,18
758
.30/
74.3
03.
00/6
.10
21.1
0/11
.70
16.9
0/7.
800.
40 (
0.01
)0.
50 (
0.01
)0.
42 (
0.01
)V
erba
l (G
ener
al
Inte
llect
ual A
bilit
y/Sp
ecifi
c A
ptitu
de
or T
alen
t)
E +
M +
HM
idw
est
Jone
s (2
010)
13,5
0017
.00/
16.5
03.
00/—
8.00
80.5
0/83
.50
1.42
74—
2.16
88V
erba
l (C
alifo
rnia
St
anda
rds
Tes
t)E
+ M
+ H
Wes
t
Ket
tler
(201
4)20
840
.00/
35.6
047
.00/
60.0
04.
00/0
.00
9.00
/4.4
00.
31 (
1.08
)0.
0000
0.47
(1.
09)
Ver
bal (
Cog
AT
)E
Sout
hwes
tLa
nge
(200
4)13
,378
75.7
0/77
.00
11.3
2/20
.60
6.90
/0.8
84.
90/1
.30
0.17
(0.
26)
0.12
(0.
41)
0.26
(0.
34)
Ver
bal (
Otis
Len
non
Abi
lity
Tes
t) +
V
erba
l (St
anfo
rd
Ach
ieve
men
t T
est)
EN
orth
Lee
and
Ols
zew
ski-
Kub
ilius
(20
06)
500,
000
80.0
0/84
.00
1.30
/7.0
014
.00/
4.00
4.70
/2.0
00.
28 (
0.03
)0.
26 (
0.03
)0.
41 (
0.04
)V
erba
l (C
alifo
rnia
A
chie
vem
ent
Tes
t)/V
erba
l (C
ompr
ehen
sive
T
est
of B
asic
Sk
ills)
/Ver
bal
(Iow
a T
est
of
Basi
c Sk
ills)
/V
erba
l (St
anfo
rd
Ach
ieve
men
t T
est)
E +
MM
idw
est
Lew
is e
t al
. (20
07)
175
58.3
0/68
.36
——
41.7
0/31
.70
0.65
(0.
36)
—0.
65 (
0.36
)N
onve
rbal
(R
AV
EN)
E +
MM
idw
est
175
58.3
0/83
.30
——
41.7
0/16
.70
0.28
(0.
79)
—0.
28 (
0.79
)V
erba
l (Io
wa
Tes
t of
Ba
sic
Skill
s)E
+ M
Mid
wes
t
175
58.3
0/86
.70
——
41.7
0/13
.30
0.21
(0.
77)
—0.
21 (
0.77
)N
onve
rbal
(N
NA
T)
E +
MM
idw
est
Lind
sey
(201
2)1,
250
48.2
0/70
.00
2.60
/—2.
90/2
.00
42.6
0/26
.00
0.47
(0.
19)
0.68
(0.
61)
0.47
(0.
19)
Ver
bal (
Cog
AT
)/V
erba
l (IT
BS)/
Ver
bal (
Stat
e A
sses
smen
t)/
Non
verb
al
(NN
AT
)
MSo
uthw
est
(con
tinue
d)
Tab
le 1
. (c
ontin
ued)
158
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Linn
-Coh
en a
nd
Her
tzog
(20
07)
500
4.00
/66.
703.
00/2
1.70
2.00
/3.3
011
.00/
5.00
0.61
(0.
49)
1.67
(0.
79)
0.43
(0.
61)
Ver
bal (
Otis
Len
non
Abi
lity
Tes
t)/
Ver
bal (
Stan
ford
A
chie
vem
ent
Tes
t)/V
erba
l (C
alifo
rnia
Tes
t of
Ba
sic
Skill
s)
EW
est
Lohm
an a
nd
Gam
brel
l (20
12)
6,97
0—
11.8
0/26
.40
40.5
0/45
.50
47.7
0/28
.10
0.37
(0.
11)
1.23
(0.
10)
0.43
(0.
11)
Ver
bal (
Cog
AT
)E
Nat
iona
l8,
261
—16
.00/
22.6
037
.20/
40.8
046
.80/
36.6
00.
65 (
0.11
)1.
16 (
0.09
)0.
66 (
0.09
)V
erba
l (C
ogA
T)
EN
atio
nal
Lohm
an e
t al
. (2
008)
1,19
834
.40/
59.3
2—
—65
.60/
40.6
80.
36 (
0.19
)—
0.36
(0.
19)
Non
verb
al (
NN
AT
)E
Sout
hwes
t1,
198
34.4
0/64
.38
——
65.6
0/33
.62
0.27
(0.
20)
—0.
27 (
0.20
)N
onve
rbal
(C
ogA
T)
ESo
uthw
est
1,19
834
.40/
57.9
8—
—65
.60/
42.0
20.
38 (
0.19
)—
0.38
(0.
19)
Non
verb
al (
RA
VEN
)E
Sout
hwes
tM
akel
, Li,
Puta
llaz,
an
d W
ai (
2011
)11
0,47
658
.47/
91.4
02.
05/2
.80
31.3
9/4.
306.
74/0
.50
0.08
(0.
06)
0.10
(0.
07)
0.07
(0.
20)
Ver
bal (
SAT
)M
Nat
iona
l
Mat
thew
s (2
006)
96,8
5858
.00/
92.2
02.
00/2
.30
31.0
0/4.
107.
00/1
.00
0.09
(0.
07)
0.10
(0.
07)
0.13
(0.
15)
Ver
bal (
Ach
ieve
men
t or
inte
llige
nce
test
)
M +
HSo
uth
McB
ee (
2006
)33
9,89
428
.70/
43.8
01.
70/5
.40
56.3
0/42
.50
13.2
0/8.
100.
45 (
0.02
)0.
03 (
0.02
)0.
25 (
0.04
)V
erba
l (st
anda
rdiz
ed
test
sco
res)
/N
onve
rbal
(T
orra
nce
Tes
ts
of C
reat
ive
Thi
nkin
g—Fi
gura
l)/V
erba
l (p
sych
omet
ric
asse
ssm
ent)
ESo
uth
Mon
rad
et a
l. (2
005)
455,
515
69.0
0/81
.00
—/1
.00
23.0
0/15
.00
6.00
/2.0
00.
50 (
0.01
)0.
59 (
0.01
)0.
32 (
0.03
)V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HSo
uth
2,67
5,02
449
.60/
63.1
72.
20/4
.23
23.9
0/9.
6123
.90/
19.5
20.
45 (
0.01
)0.
34 (
0.01
)0.
77 (
0.01
)V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HSo
uth
719,
000
59.0
0/80
.57
2.00
/—35
.00/
15.7
65.
00/—
0.28
(0.
01)
0.35
(0.
01)
—V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HSo
uth
1,55
9,82
848
.00/
74.8
63.
00/5
.55
38.0
0/15
.21
8.00
/2.2
00.
25 (
0.01
)0.
29 (
0.01
)0.
26 (
0.02
)V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HSo
uth
1,32
5,70
758
.47/
83.7
82.
05/3
.16
31.3
9/10
.45
6.74
/1.8
20.
23 (
0.01
)0.
26 (
0.01
)0.
27 (
0.02
)V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HN
orth
1,20
3,69
759
.75/
76.0
44.
80/8
.49
26.7
0/10
.51
7.00
/3.2
20.
31 (
0.01
)0.
32 (
0.01
)0.
44 (
0.02
)V
erba
l (A
chie
vem
ent
or A
ptitu
de t
est)
E +
M +
HSo
uth
Nag
lieri
and
For
d (2
003)
18,9
9574
.40/
79.5
4—
15.0
7/12
.73
10.4
8/7.
730.
75 (
0.08
)0.
86 (
0.09
)0.
87 (
0.11
)N
onve
rbal
(N
NA
T)
E +
M +
HN
atio
nal
(con
tinue
d)
Tab
le 1
. (c
ontin
ued)
159
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Neu
mei
ster
, Ada
ms,
Pi
erce
, Cas
sady
, an
d D
ixon
(20
07)
3,44
331
.00/
49.5
0—
58.0
0/30
.50
9.00
/14.
000.
39 (
0.11
)0.
32 (
0.12
)1.
65 (
0.17
)V
erba
l (T
erra
Nov
a C
ompr
ehen
sive
T
est
of B
asic
Sk
ills)
/Non
verb
al
(RA
VEN
)/N
onve
rbal
(A
dam
s/Pi
erce
C
heck
list)
EM
idw
est
Offi
ce fo
r C
ivil
Rig
hts
(201
5)48
,273
,920
54.9
4/65
.00
5.17
/9.5
816
.70/
9.86
22.2
6/15
.44
0.53
(0.
00)
0.55
(0.
00)
0.64
(0.
00)
——
Nat
iona
l
Ols
zew
ski-K
ubili
us
and
Tur
ner
(200
2)
70,0
5859
.00/
83.0
03.
00/5
.00
21.0
0/2.
4017
.00/
1.00
0.06
(0.
07)
0.09
(0.
09)
0.05
(0.
14)
Ver
bal
(Ach
ieve
men
t, A
ptitu
de, o
r A
bilit
y te
st)
EM
idw
est
Ols
zew
ski-K
ubili
us,
Lee,
Ngo
i, an
d N
goi (
2004
)
3,23
845
.60/
95.0
03.
00/—
43.7
0/5.
007.
10/—
0.05
(0.
36)
0.07
(0.
36)
—N
onve
rbal
(N
NA
T)/
Ver
bal
(Ach
ieve
men
t te
st)
EM
idw
est
3,23
845
.60/
89.0
03.
00/—
43.7
0/11
.00
7.10
/—0.
12 (
0.25
)0.
16 (
0.25
)—
Non
verb
al
(NN
AT
)/ V
erba
l (A
chie
vem
ent
test
)
EM
idw
est
3,23
845
.60/
92.0
03.
00/—
43.7
0/8.
007.
10/—
0.08
(0.
29)
0.11
(0.
29)
—N
onve
rbal
(N
NA
T)/
Ver
bal
(Ach
ieve
men
t te
st)
EM
idw
est
Peai
rs, E
iche
n,
Puta
llaz,
C
osta
nzo,
and
G
rim
es (
2011
)
474
40.0
0/70
.00
—46
.00/
27.0
0—
0.43
(0.
33)
0.43
(0.
33)
—V
erba
l (A
chie
vem
ent
or a
ptitu
de t
est)
MN
orth
Pend
arvi
s an
d W
ood
(200
9)4,
800
89.0
0/90
.00
—11
.00/
10.0
0—
0.90
(0.
16)
0.90
(0.
16)
—V
erba
l (W
ISC
)E
+ M
Sout
h
Pete
rs a
nd G
entr
y (2
012)
464
38.3
0/71
.00
—61
.70/
29.0
0—
0.25
(0.
32)
0.25
(0.
32)
—V
erba
l (st
anda
rdiz
ed)
E +
MM
idw
est
539
57.0
0/71
.00
—43
.00/
29.0
0—
0.54
(0.
31)
0.54
(0.
31)
—V
erba
l (st
anda
rdiz
ed-
loca
l nor
ms)
E +
MM
idw
est
539
57.0
0/62
.50
—43
.00/
37.5
0—
0.80
(0.
19)
0.80
(0.
19)
—V
erba
l (st
anda
rdiz
ed-
loca
l nor
ms)
E +
MM
idw
est
539
57.0
0/68
.00
—43
.00/
32.0
0—
0.62
(0.
22)
0.62
(0.
22)
—V
erba
l (st
anda
rdiz
ed-
loca
l nor
ms)
E +
MM
idw
est
(con
tinue
d)
Tab
le 1
. (c
ontin
ued)
160
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Pfei
ffer
and
Jaro
sew
ich
(200
7)59
264
.00/
63.0
04.
00/6
.00
16.0
0/16
.00
16.0
0/15
.00
0.95
(0.
29)
1.00
(0.
37)
0.93
(0.
38)
Non
verb
al (
Rat
ing
Scal
es)
EN
atio
nal
Pfei
ffer,
Pet
sche
r,
and
Jaro
sew
ich
(200
7)
395
62.6
7/70
.80
2.67
/4.4
416
.00/
13.1
616
.00/
9.91
0.64
(0.
24)
0.80
(0.
31)
0.58
(0.
34)
Non
verb
al (
Rat
ing
Scal
es)
E +
MN
atio
nal
Pier
ce e
t al
. (20
06)
3,58
43.
00/4
6.00
—58
.00/
36.0
09.
00/1
2.70
0.47
(0.
12)
0.41
(0.
12)
1.47
(0.
18)
Non
verb
al (
RA
VEN
)E
Mid
wes
tPi
erce
et
al. (
2011
)75
027
.00/
30.0
0—
53.0
0/33
.00
14.0
0/28
.00
0.77
(0.
18)
0.44
(0.
18)
2.39
(0.
21)
Non
verb
al (
RA
VEN
)E
Mid
wes
t65
025
.00/
43.0
0—
55.0
0/32
.00
14.0
0/16
.00
0.41
(0.
20)
0.39
(0.
20)
1.17
(0.
27)
Non
verb
al (
RA
VEN
)E
Mid
wes
tR
atcl
iff e
t al
. (20
12)
587
54.0
0/90
.00
1.00
/—41
.00/
6.67
1.00
/3.3
30.
15 (
0.44
)0.
10 (
0.52
)3.
41 (
0.83
)V
erba
l (A
chie
vem
ent
test
)H
Sout
h
Rob
erts
on (
2013
)44
,938
60.0
0/75
.00
4.80
/5.0
026
.70/
10.0
07.
30/4
.00
0.32
(0.
09)
0.31
(0.
19)
0.53
(0.
30)
Ver
bal (
Ach
ieve
men
t te
st)
HSo
uth
Row
e, M
iller
, Eb
enst
ein,
and
T
hom
pson
(20
12)
1,68
046
.00/
46.0
044
.00/
45.0
0—
2.00
/4.0
02.
04 (
0.43
)—
2.04
(0.
43)
Ver
bal (
WIS
C)
ESo
uth
Saro
uphi
m (
2002
)30
320
.80/
14.6
0—
—50
.20/
57.3
01.
33 (
0.31
)—
1.33
(0.
23)
Non
verb
al
(per
form
ance
ta
sks)
HSo
uthw
est
Saro
uphi
m (
2004
)39
514
.70/
12.3
0—
—38
.20/
40.8
01.
11 (
0.44
)—
1.11
(0.
30)
Non
verb
al
(per
form
ance
ta
sks)
MSo
uthw
est
Scot
t an
d D
elga
do
(200
5)26
221
.00/
77.0
0—
40.0
0/4.
0039
.00/
19.0
00.
08 (
0.47
)0.
06 (
1.01
)0.
37 (
0.51
)V
erba
l (IQ
)E
Sout
h
Shau
ness
y-D
edri
ck,
Evan
s, F
erro
n,
and
Lind
o (2
015)
194,
000
40.0
0/75
.20
3.00
/4.0
021
.00/
9.00
29.0
0/1.
800.
12 (
0.03
)0.
37 (
0.03
)0.
04 (
0.06
)V
erba
l (IQ
)E
Sout
h
Suld
o an
d Sh
aune
ssy-
Ded
rick
(20
13)
1,67
561
.85/
64.9
32.
52/3
.73
16.3
1/9.
8218
.48/
9.82
0.46
(0.
22)
0.56
(0.
30)
0.48
(0.
30)
Ver
bal (
Stat
e ac
hiev
emen
t te
st)
HSo
uth
Swia
tek
and
Lupk
owsk
i-Sh
oplik
(20
03)
117,
215
84.1
0/84
.70
1.80
/4.1
010
.00/
1.30
3.20
/0.2
00.
10 (
0.12
)0.
12 (
0.13
)0.
06 (
0.33
)V
erba
l (St
ate
achi
evem
ent
test
)E
Nor
th
Ter
ry (
2003
)77
371
.40/
88.0
00.
40/2
.00
27.7
0/10
.00
0.50
/—0.
28 (
0.39
)0.
29 (
0.39
)—
—M
Sout
hT
exas
Edu
catio
n A
genc
y (2
011)
648,
111
33.7
0/37
.00
3.50
/11.
2013
.10/
7.10
47.4
0/41
.95
0.63
(0.
02)
0.51
(0.
02)
0.80
(0.
01)
——
Sout
hwes
t
(con
tinue
d)
Tab
le 1
. (c
ontin
ued)
161
Aut
hors
(ye
ar)
Size
(N
)G
P/ID
W %
GP/
ID A
%G
P/ID
B %
GP/
ID H
%O
vera
ll ri
sk
ratio
(SE
)R
isk
ratio
B
(SE)
Ris
k R
atio
H
(SE)
ID m
etho
dsG
rade
Reg
ion
Tex
as E
duca
tion
Age
ncy
(201
5)41
7,49
01.
65/3
.20
0.47
/1.4
80.
18/0
.15
97.5
6/94
.95
0.45
(0.
03)
0.83
(0.
14)
0.47
(0.
02)
——
Sout
hwes
t42
2,50
91.
59/2
.97
0.45
/1.5
20.
19/0
.17
97.6
3/95
.14
0.45
(0.
03)
0.89
(0.
13)
0.48
(0.
03)
——
Sout
hwes
t42
3,92
11.
57/2
.66
0.44
/1.5
40.
23/0
.15
97.6
0/95
.43
0.48
(0.
03)
0.65
(0.
14)
0.51
(0.
03)
——
Sout
hwes
t49
,879
1.00
/2.1
30.
27/0
.84
0.09
/0.1
798
.61/
96.8
70.
43 (
0.09
)1.
89 (
0.40
)0.
44 (
0.09
)—
—So
uthw
est
49,1
900.
98/1
.39
0.27
/0.7
00.
09/0
.06
98.6
4/97
.85
0.60
(0.
10)
0.67
(0.
61)
0.63
(0.
10)
——
Sout
hwes
t49
,370
0.80
/1.1
00.
20/0
.52
0.07
/0.0
698
.90/
98.3
20.
63 (
0.11
)0.
86 (
0.60
)0.
65 (
0.11
)—
—So
uthw
est
48,3
550.
62/0
.81
0.19
/0.5
00.
10/0
.08
99.0
5/98
.59
0.64
(0.
13)
0.80
(0.
52)
0.67
(0.
12)
——
Sout
hwes
tV
an T
asse
l-Bas
ka,
Feng
, and
De
Brux
(20
07)
140,
000
59.0
0/82
.90
1.00
/2.4
035
.00/
11.8
05.
00/1
.60
0.23
(0.
02)
0.25
(0.
02)
0.31
(0.
05)
Ver
bal (
Stat
e ac
hiev
emen
t te
st)
ESo
uth
140,
000
59.0
0/80
.60
1.00
/2.1
035
.00/
14.0
05.
00/2
.10
0.29
(0.
03)
0.30
(0.
03)
0.41
(0.
08)
Non
verb
al
(per
form
ance
ta
sks
with
m
anip
ulat
ives
)
ESo
uth
Van
Tas
sel-B
aska
, Fe
ng, a
nd E
vans
(2
007)
139,
611
54.0
0/85
.70
1.00
/2.2
041
.00/
9.80
1.00
/1.6
00.
18 (
0.03
)0.
16 (
0.03
)1.
61 (
0.08
)V
erba
l (St
ate
achi
evem
ent
test
)E
Sout
h
139,
611
54.0
0/84
.57
1.00
/2.1
141
.00/
11.1
01.
00/1
.50
0.20
(0.
02)
0.18
(0.
03)
1.51
(0.
07)
Non
verb
al
(per
form
ance
ta
sks
with
m
anip
ulat
ives
)
ESo
uth
Van
Tas
sel-B
aska
, Jo
hnso
n, a
nd
Ave
ry (
2002
)
716,
769
46.2
0/80
.00
1.30
/—42
.10/
15.0
02.
40/—
0.22
(0.
01)
0.24
(0.
01)
——
—So
uth
1,79
268
.00/
83.3
0—
23.0
0/11
.60
9.00
/5.0
00.
42 (
0.13
)0.
44 (
0.15
)0.
53 (
0.22
)N
onve
rbal
(p
erfo
rman
ce
task
s w
ith
man
ipul
ativ
es)
ESo
uth
War
ne e
t al
. (20
13)
13,9
3880
.50/
78.5
01.
30/3
.60
—18
.20/
17.9
00.
98 (
0.03
)—
0.98
(0.
03)
—E
Wes
tW
ilson
(20
15)
3494
.10/
93.3
02.
90/6
.70
2.90
/0.0
0—
0.00
000.
0000
Non
verb
al (
Prim
ary
test
of N
onve
rbal
In
telli
genc
e)
ESo
uthw
est
Woi
tasz
ewsk
i and
A
alsm
a (2
004)
631
80.7
0/87
.00
1.90
/3.0
09.
50/5
.00
6.40
/3.0
00.
46 (
0.60
)0.
50 (
0.75
)0.
45 (
0.95
)V
erba
l (W
ISC
)H
Mid
wes
t
Not
e. G
P/ID
= G
ener
al P
opul
atio
n/Id
entif
ied
Popu
latio
n; W
= W
hite
; A =
Asi
an; B
= B
lack
; H =
His
pani
c; E
= E
lem
enta
ry S
choo
l; H
= H
igh
Scho
ol; M
= M
iddl
e Sc
hool
; SE
= s
tand
ard
erro
r; W
ISC
= W
echs
ler
Inte
llige
nce
Scal
e fo
r C
hild
ren;
ITBS
= Io
wa
Tes
t of
Bas
ic S
kills
; Cog
AT
= C
ogni
tive
Abi
litie
s T
est;
SAT
= S
chol
astic
Ass
essm
ent
Tes
t; N
NA
T =
Nag
lieri
Non
verb
al A
bilit
ies
Tes
t; R
AV
EN =
Rav
en S
tand
ard
Prog
ress
ive
Mat
rice
s.
Tab
le 1
. (c
ontin
ued)
162 Gifted Child Quarterly 62(2)
Outliers. Before inferential analysis was conducted, the dis-tribution of the RRs was examined for any outliers. Outliers should be eliminated when illegitimately present in a data set; but their exclusion should be approached with rigor and caution (Barnett & Lewis, 1994). Tukey (1960, 1977) espoused elimination of outliers that exist three times the interquartile range of either the 75th or 25th percentiles. One study fit this criterion. Wilson (2015) conducted a study observing differ-ences between young children with high intellectual abilities and their peers in cognitive and social play. A sample-adjusted meta-analytic deviancy statistic (Huffcutt & Arthur, 1995) was calculated for the study and indicated that inclu-sion or exclusion of the study would have minimal influence on estimating the population parameters due to the small pri-mary sample size (n = 34) in the Wilson (2015) study. As the study was minimally impactful, it was only included in anal-ysis for the overall effect size and was excluded from subse-quent analysis.
Data Analysis
A random-effects model was used as the theoretical frame-work for conducting a meta-analysis. Given that these effect sizes are associated with social science data, no perfect mea-sure can be derived to accurately formulate the parameter that describes the population of analysis, and thus some of the variation might be simply due to error. However, we view the population value for the effect sizes retrieved from pri-mary studies as not identical but follow a hypothetical distri-bution. Thus, the observed variation in effect sizes is also caused by differences of design in the studies, in particular by different methods of identification and student population in primary studies.
The Computation of Average RR and Exploration of Variance in RRs. We computed the average RR to determine if Black, Hispanic, and/or Native American students are proportion-ally represented in gifted programs. In order to compute the average RR, we first transformed RR to log risk ratio (LRR) by taking the natural logarithm of RR to maintain a symmet-ric distribution of effect sizes (Borenstein et al., 2009). The SE of the LRR is defined as:
SEURG UR RG RLRRi
i i i i
= − + −1 1 1 1
where URi is the number of Black, Hispanic, and/or Native
American students in study i, URGi is the number of Black,
Hispanic, and/or Native American students who were identi-fied as gifted among URG
i, R
i is the number of White and
Asian students in the same study, and RGi is the number of
White and Asian students who are identified among Ri. We
used the weight, which is the inverse of total variance, to compute the average LRR, so that the primary study with large sample size will obtain larger weight than that with small sample size. We ran Q test of heterogeneity to judge if
observed variance is beyond sampling errors. Second, we used I2 (Higgins & Thompson, 2002) and τ 2 indices to understand the extent of the variance resulted in between-study difference relative to total variation in LRR.
We repeated the aforementioned analyses for (a) the com-bined group of traditionally underrepresented students (i.e., Black, Hispanic, and Native American students) compared with the combined group of White and Asian students and (b) each racial minority group compared separately with the combined group of White and Asian students. Note that we could not conduct a subgroup analysis for Native American students because of the small number of effect sizes (k = 9). All results of inferential analyses were transformed back to the original metric and reported in the Results section.
Moderator Analysis
Three factors were primarily explored in the moderator analyses to explain the variation in RRs: (1) type of identification meth-ods (five methods), (2) geographic locations (four regions), and (3) grade level (three grade levels). To address the second research question, we considered analyzing all three categorical moderators simultaneously with a meta-regression. However, conducting the regression analysis resulted analyzing nine dummy variables in the model using a reduced number of effect sizes (k = 30). The reduction of the sample size occurs because not all studies include information for all variables. For exam-ple, in the study by Lohman et al. (2008), the authors only pro-vided information on Hispanic and White student populations but did not include any information about the Asian and Black students. As a result, this study would not be included in the regression analysis. Eliminating studies that did not provide information on all variables in regression analysis for listwise deletion reduced the sample size, which would result in lower-ing the statistical power and affecting the precision and accu-racy of the estimations. In addition, although regression analysis allows researchers to examine an additive effect of each mod-erator when controlling other moderators in the model, while the interpretation of those coefficients become more complex. In fact, we found that the three moderators were relatively inde-pendent in our preliminary analyses (i.e., |r
IDLocation| = .07-.33,
|rIDGrade
| = .01-.16, |rLocationGrade
| = .004-.14, all are nonsignificant, except for IQ and midwest), suggesting that no specific applica-tion of identification method is used at a particular region or grade level and thus the influence of intercorrelation among moderators was minimal. Furthermore, Higgins and Thompson (2002) indicated that a meta-regression with multiple covariates (in our case, nine dummy variables with one intercept) will increase the likelihood of committing a Type I error when evalu-ating individual effects. Based on methodological consideration for the current data conditions and for ease of interpretation, we chose to use meta–analysis of variance (ANOVA) to analyze each moderator separately. In addition, due to the increased pos-sibility of committing a Type I error with multiple ANOVA, we adjusted alpha level to α = .01 for statistical evaluation.
Hodges et al. 163
Comprehensive Meta-Analysis software (Version 3.1.1; Borenstein, Hedges, Higgins, & Rothstein, 2014) was used for the analyses using meta-ANOVA.
Sensitivity Analysis
Publication Bias. Since studies with contradictory or null results are less likely to be published than studies with significant results in the expected direction, description of unpublished studies is required. Among the studies in the current meta-analysis, 45 of them are published, 4 of them are unpublished dissertations, and the remaining 5 are state and technical reports. We constructed a funnel plot to describe the distribution of effect sizes in terms publication bias (Figure 2). The idea behind the funnel plot is that if the effect sizes are distributed symmetri-cally in a funnel shape (with the idea that the larger the sample size, the closer to the true population mean) then there can be assumed to be little publication bias (Cooper, Hedges, & Valen-tine, 2009). Analysis of the funnel plot showed minimal publi-cation bias as indicated by the roughly symmetrical shape. A classic fail-safe analysis was conducted as an alternative test for publication bias and indicated that 4,113,553 studies would be needed for p > .05 to be observed. Orwin’s fail-safe was also used to test how many studies would need to be added to the meta-analysis for the null to be accepted (RR = 1). The test results indicated 6,115,182 null studies would need to be added for the observed effect with the current pool of studies for the null result to be accepted.
Results
Representation of Black, Hispanic, and Native American Students in Gifted Program
The overall average RR, that compares the proportion of Black, Hispanic, and Native American students identified as gifted and/or served in a gifted program to that of the
reference group, was 0.34 with an SE of 0.01. The RR of 0.34 indicates that the probability of being identified for Black, Hispanic, and Native American students is about one-third that of the probability of being identified for White and Asian students. This means that students in the focal group remain largely underidentified for and underrepresented in gifted programs in the United States. However, as shown the variability of RRs in Table 2, the Q result also indicated that the significant variation exists among retrieved RR, Q(84) = 149,293.87, p < .0, τ2 = .081. Because the I2 index of 99.99 % indicates heterogeneity of effect sizes are mainly due between study variance, this supports the need of moderator analyses (Hox & Leeuw, 2003).
Moderator Analyses. Because not all studies reported identifi-cation methods, only 56 of 85 effect sizes were associated with a description of the identification procedures used in their respective studies. For example, a study only contain-ing information on grade level would only be included in the meta ANOVA regarding grade level. A test for heterogeneity supports that significant variability across effect sizes existed, Q(55) = 8,277.60, I2 = 99.08, p < .01, τ2 = 0.09. As reported in Table 2, the average RR for nonverbal identifica-tion methods is 0.34 (k = 28, SE = 0.08); whereas, for 30 effect sizes were aggregated to provide an RR effect size of 0.27 (SE = 0.06) for the traditional methods. However, the difference in the average RRs was not statistically signifi-cant, Q(1) = 2.31, p > .50. The inclusion of additional effect sizes in the meta ANOVA breakdown by a specific identifi-cation instrument also suggested that the difference is not statistically significant, Q(4) = 3.21, p > .50. The average RR was 0.42 for RAVEN (k = 7, SE = 0.26) while the average RRs were 0.27 for NNAT (k = 11, SE = 0.12) and 0.34 (k = 4, SE = 0.26) for the CogAT-NV, respectively. Conversely, the average RR for IQ tests was 0.31 (k = 10, SE = 0.16) and that of achievement tests was 0.24 (k = 20, SE = 0.08). This sug-gests that regardless of the identification instruments, the
Figure 2. Funnel plot for publication bias.
164 Gifted Child Quarterly 62(2)
Figure 3. Forest plot for examining effect sizes.
Hodges et al. 165
probability of being identified for the underrepresented group is persistently about one-third of that for the overrep-resented group. Similarly, the chance of being identified is persistently lower across all grade levels for underrepre-sented groups compared with overrepresented group, Q(2) = 3.63, p = .16, in the meta ANOVA.
An examination of the effect sizes for location suggested that there existed a difference in identification rate across location, Q(3) = 20.07, I2 = 99.74, p < .001 (see Table 2). We found the average RR of 0.47 for southwest (k = 19, SE = 0.09) and 0.30 for the south (k = 22, SE = 0.06), while the RRs for north and midwest are 0.23 (k = 5, SE = 0.17) and 0.24 (k = 24, SE = 0.04), respectively. This indicates that the southwest had the highest rates of proportional identification with underrepresented populations being identified at nearly 50% of the rate of populations who are traditionally identi-fied for services.
Representation of Black Students in Gifted Education Program
An overall RR was 0.28 (k = 74, SE = 0.03) with significant variation in RR across studies, Q(73) = 122,697.08, p < .001, I2 = 99.90. The subsequent moderator analyses identified patterns like those we observed in the results with overall underrepresented population, except participant grade level in the meta-ANOVA. More specifically, the identification bias is persistent regardless of identification instruments.
In addition, in the meta-ANOVA with only Black students in the focus group, we found that the RR was higher (RR = 0.47, k = 9, SE = 0.25) when studies combined elementary and middle school students in their sample, compared with elementary only (RR = 0.17, k = 31, SE = 0.08). We also
found strong moderator effect by location, Q(3) = 17.25, p < .001. The probability of being identified for Black students compared with White and Asian students is less than one-fifth that of the midwest (i.e., RR = 0.17, k = 24, SE = 0.03). However, the probability of being identified in southwest is greater (RR = 0.46, SE = 0.10), but still about 50% less likely compared with the reference group.
Representation of Hispanic Students in Gifted Education Program
The moderator analyses for Hispanic students showed a somewhat different pattern from than that for Black students. An overall RR is 0.36 (SE = 0.03) with significant variation across studies, Q(71) = 82,541.03, p < .001, I2 = 99.12, which is similar to the overall RR for Black students. Although it was not statistically significant with the specified alpha of .01, Q(1) = 4.13, p = .04, possibly large variation in RRs within the category, it is worth noting that the RR for identi-fication by nonverbal tests was 0.50 (k = 22, SE = 0.26), whereas, it was only 0.26 (k = 27, SE = 0.12) for verbal iden-tification measures (see Table 3). Furthermore, inconsistent with the result of Black students, there is no difference in RRs by grade level, Q(2) = 0.88, p >.50. Like the results with Black students, we found no significant differences in RR among specific identification tests we explored, Q(4) = 1.36, p >.50. We also found that significant variation in the aver-age RR by geographical location for Hispanic students. Similar to the finding with Black students, identification bias is less in the south (RR = 0.34, k = 18, SE = 0.10), and south-west (RR = 0.49, k = 19, SE = 0.13), Q(3) = 13.27, p = .01. Thus, the probability of being identified for Hispanic stu-dents living in in south and southwest is greater than for
Table 2. The Average Risk Ratio (RR) as a Function of Type of Identification Test, Study Participants’ Grade Level, and Location.
those students in the same racial group who live in other region in the United States, but still about 50% less likely compared with the reference group.
Discussion
Scholars in the field of gifted education have observed dis-crepancies in racial representation in gifted programming (Yoon & Gentry, 2009) as well as in identification by tradi-tional testing methods, including state standardized tests and IQ tests (D. Ford, 2014). With an overall RR of 0.34, this means that the probability of historically underrepresented students being identified is 66% less than for Asian and White students. This supports previous findings that Black and Hispanic students continue being largely underidentified and underrepresented in gifted programs in the United States for decades (Daniels, 1998; Erwin & Worrell, 2012; D. Ford, 2014; Yoon & Gentry, 2009). However, when this RR is examined using a moderator of location or grade, the results indicate a more positive outlook for some groups in certain regions.
Black students were identified with an RR of 0.28, evidenced their underrepresentation in gifted programs. This finding is aligned with the body of research that details the inequity in representation reported by numerous researchers (e.g., Daniels, 1998; Erwin & Worrell, 2012; D. Ford, 2014; D. Y. Ford & Harris, 1994; Lakin & Lohman, 2011; Naglieri & Ford, 2003; Yoon & Gentry, 2009). Additionally, when this national rate is examined by region, the midwest woefully underidentifies
Black students (RR = 0.17). In comparison, the south identi-fies Black students at almost twice the rate of the midwest (RR = 0.29) and the southwest identifies them at almost three times the rate of the midwest (RR = 0.46). The identification gap observed by many researchers for Black students exists across throughout the United States. Warne et al. (2013) reported an OR of identification of Black students (OR = 0.81), Hispanic (OR = 0.95), and Native American (OR = 0.54) students in Utah. An OR of 0.81 meant that the Black students are being identified at 81% the rate of identification when compared with their White peers. The researchers used their findings to state that there was no unified underrepre-sentation across the United States for traditionally underrep-resented populations due to the comparatively high rates of identification in Utah. In comparison, the identification sta-tistic calculated for Black students in our study is closer to the one observed by Konstantopoulos et al. (2001), in which they reported an OR of 0.37 for Black students in compari-son with their White peers. Given that equitable identifica-tion of Black students was not observed in other authors’ studies, it is likely that the findings of Warne et al. (2013) are unique to the state of Utah. As such, these findings supported the conclusion drawn by the authors that there was variabil-ity in identification by region. These findings also support the work of Yoon and Gentry (2009), who documented vari-ability by state in identification rates by using the OCR reporting database. Some explanations for the uneven identi-fication rate could be due to the litigation leveled within some southern states in particular (Stephen et al., 2012),
Table 3. Risk Ratio (RR) of the Various Moderators for Black students and Hispanic Students.
which has led to changes in policy and resulting identifica-tion practices.
With an overall RR of 0.36, Hispanic students, like Black students, are underrepresented in gifted programs. In addition, when examined by region, the south (RR = 0.34) and south-west (RR = 0.49) have higher levels of proportional identifica-tion than other regions, particularly the midwest (RR = 0.29). Esquierdo and Arreguín-Anderson (2012) voiced their con-cern over the lack of proportional identification in the south-west among Hispanic students, but compared with the national trend, these students are identified as gifted at 33% greater in this region than in other areas of the country. Using the data presented by Esquierdo and Arreguín-Anderson (2012), an RR of 0.48 was reported for Hispanic students in the south-west. This coincides with the RR found in our study. This higher level of proportional identification suggests that the education systems in the southwest are having success in the methods and policies it uses for identification.
There is also a likely corollary between districts with large numbers of Hispanic students and higher identification rates (Raudenbush et al., 1998). Consequently, the midwest, which is currently experiencing an increase in enrollment in public schools by Hispanic students (Brown & Lopez, 2013), is below the national average for proportional identification of Hispanic students. This supports the findings of Yoon and Gentry (2009). In the south and southwest regions of the country, despite the increase in identification of Hispanic students, they are still largely underrepresented in gifted programs.
In comparing the identification rate between Black and Hispanic students, Black students are identified at lower rates than their Hispanic peers in all regions examined. The greatest contrast between the two rates in comparison with the overall rate is in the midwest (Black RR = 0.17, Hispanic RR = 0.29). This is a new finding in the field. Scholars who have examined underrepresentation have largely focused on the south and southwest regions of the United States (Esquierdo & Arreguín-Anderson, 2012; McBee, 2006). As such, a large gap in proportional identification existing in the midwest among Black students and their Hispanic peers is a finding that warrants further investigation.
Statistically significant differences in terms of grade level were only observed for Black students. These students were identified at higher rates in the middle grades (RR = 0.47) compared with the elementary (RR = 0.17) and high school (RR = 0.18) levels. Consequently, this suggests that the trend of closing performance gaps on tests at the later elementary and the middle level observed by Lohman and Korb (2006) may have led to higher identification rates. The large SE associated with the effect size for middle grades suggests a large amount of variance among identification rates of mid-dle school Black students. However, Black students may have been less likely to be identified as gifted because many enter public school with less academic exposure than other students (D. Y. Ford, Grantham, & Whiting, 2008). These
students “catch up” to their peers by late elementary and middle school and are therefore more likely to be identified as gifted at that time.
The moderator analysis of identification methods pro-vided notable empirical evidence contradicting the claims of researchers who have developed nonverbal methods as a means to close the proportional representation gap (Lohman, 2011; Naglieri & Ford, 2003; Pfeiffer & Jarosewich, 2007; Sarouphim, 2002; Van Tassel-Baska et al., 2002). Overall, we found no statistically significant difference in the RR between verbal and nonverbal methods of identification. In addition, no statistically significant differences existed for the testing meth-ods examined in this meta-analysis. The nonverbal tests were specifically created to address the proportional gap in identi-fication of underserved students (Naglieri & Ford, 2003), but the gap persists.
An examination of the influence of verbal and nonverbal methods of identification in Black and Hispanic students, separately, revealed a difference in the average RRs for the two groups. Black students were not only being identified at the same rate with regard to verbal and nonverbal identifi-cation methods, but both rates were also remarkably low (RR = 0.17 and RR = 0.19, respectively). These rates of iden-tification demonstrate that despite researchers’ best efforts (Lohman, 2011; Naglieri & Ford, 2003; Pfeiffer & Jarosewich, 2007; Sarouphim, 2002; Van Tassel-Baska et al., 2002), Black students remain underidentified and underserved regardless of identification method (Erwin & Worrell, 2012; Yoon & Gentry, 2009). This gap in representation is unlikely to be closed by only a test without further support in the form of programs and funding (Kettler et al., 2015; Peters et al., 2013)
The findings for Hispanic students were the same as those for Black students. There was no statistically significant dif-ference between verbal and nonverbal identification meth-ods. This provides evidence that Hispanic students also experience no increase in proportional identification rates when nonverbal tests are used in place of traditional identifi-cation methods. As the Hispanic population continues to grow in public schools (Brown & Lopez, 2013), the results of the meta-analysis show that there has not been similar growth in gifted identification rates as well. Esquierdo and Arreguín-Anderson (2012) cautioned against the prospect of excluding Hispanic students from gifted services. The results from this study demonstrate that nontraditional methods of identification are unlikely to address the concerns of Esquierdo and Arreguín-Anderson (2012).
The results of this study coincide with Erwin and Worrell’s (2012) findings but not with their conclusion. In their analy-sis of assessment practices, they found that, when controlling for achievement, there was no difference in identification rates. Like Warne et al. (2013), they argued that a combina-tion of socioeconomic factors was the cause for underidenti-fication and not the actual identification methods. This conclusion is supported by the results for Black and Hispanic
168 Gifted Child Quarterly 62(2)
students for whom there was no significant difference between the verbal and nonverbal tests.
Because of the limited number of effect sizes (k = 15), we were unable to conduct similar analyses for Native American students. Of the 10 retrieved studies, seven were from state reports and three were from national sur-vey data. An overall RR effect size of 0.49 was obtained from the available data with an SE of 0.11. The lack of usable data found through this meta-analysis of the Native American students is discouraging in many ways. First, this is reflective of how researchers generally do not report on this population when examining gifted students in schools (Gentry, Fugate, Wu, & Castellano, 2014). This is troubling as the national information derived from the state reports and national surveys indicates that Native American students are underrepresented in gifted pro-gramming. However, we do not have information on how underrepresented they are in comparison with Black and Hispanic students, nor do we know if the RR for the Native American students is similar to that of the different regions of the country or among different grade levels. Second, this lack of information makes it difficult for all stakeholders to implement policies to address representation of Native youth in gifted programs. In this meta-analysis, we found the use of (non)verbal methods were able to iden-tify almost twice as many Hispanic students for gifted programming than the use of verbal methods. However, the same cannot be said for Native American students as there is not enough information about their identification rates on the various methods of scholastic assessments. This makes it difficult to understand baseline or to mea-sure improvement concerning proportional representation of Native American students.
A final finding is that articles published in gifted educa-tion journals lack a standardized reporting procedure for their samples. Of the 1,526 articles reviewed in the four major gifted education journals, the methods for reporting samples varied. Some authors reported a sample as being gifted without any description of how those students were identified as gifted. Given the broad spectrum of definitions for giftedness and the varied identification methods (Erwin & Worrell, 2012), trying to draw cohesive inferences about the population is problematic.
In essence, making inferences on a population that is ill-defined frequently leads to error and/or bias. Two studies might examine gifted populations, but if those two studies drew on samples that were identified using different methods and under various definitions, then trying to generalize between the two samples becomes difficult if not outright impossible. In other words, the term “identified gifted” has a large amount of variance in what it means. When analyzing the effects of given interventions and identification methods, researchers should provide greater clarity concerning the demographics of their sample and the identification proce-dures used to determine giftedness.
Limitations
The study has several limitations that affect the generalizability of the results. First, a relatively small number of effect sizes were used for analysis of the three specific identification tests (NNAT, RAVEN, and CogAT-NV) as compared with the other moderators (grade and region). This limitation is mitigated by the small SEs due to the large sample size within the primary studies, but with an increased number of effect sizes from large samples, the validity of the results will increase. Furthermore, the small number of effect sizes limits the scope of moderator analyses. Our analysis could investigate three key moderators. Because these three moderators are little correlated, our conclu-sion with meta-ANOVA maintains a statistical conclusion valid-ity. However, the fact that unexplained variances remain among effect sizes suggests the possibility for the existence of other moderators which are not revealed in the current analysis. There are many other factors that can influence proportional identifi-cation. Factors such as the social–economic status of the dis-tricts, the effects of poverty and immigration (e.g., language challenges, cultural differences), as well as systematic racism are likely to contribute to the underrepresentation of Black, Hispanic, and Native American students receiving gifted ser-vices. If a newly found moderator is related to one of three mod-erators used in current investigation, the analysis with meta-ANOVA does not identify the unique impact of the mod-erators without accounting for the correlated moderators. Thus, further investigation of other potential moderators will not only require additional k but also require the use of more sophisti-cated statistical analysis, such as the application of multilevel modeling (Hedges & Maier, 2013) to fully entangle the com-plex phenomena of identification gap in gifted program. It is also important to mention that further investigation is needed to facilitate our understanding of the roles these factors play as well as the strength of their influences on identification in pri-mary studies.
Second, because of the lack of consistency in reporting practices, some effect sizes have been impossible to calculate. Unless researchers studying gifted students are clear about the demographics of their sample and how the students were iden-tified as gifted, it will be difficult to obtain the necessary effect sizes needed to increase the validity of the results.
Third, the samples from the northeast and western regions of the United States were not included in a moderator analy-sis. This is partially due to the limited number of studies that have been conducted in these two regions with gifted popula-tions. Since we found that identification gaps differ by region, without information from the northeast and western regions of the country, we do not know if these regions are achieving proportional representation in their gifted popula-tions. The field of gifted education will benefit from high-quality studies conducted on the gifted population in these regions with detailed reporting that clearly explains the sam-ple demographics along with the methods used to identify the gifted students.
Hodges et al. 169
Finally, we were unable to conduct analyses for Native American students due to a lack of studies that report statis-tics for this population. The Native American population is about 1.2% of the total population of the United States. Although this is a small part of the population, it does not mean that Native Americans should not receive the same attention that the Black and Hispanic populations receive from researchers (Gentry et al., 2014). Gifted students can (and should) be found in all racial groups, and they deserve a chance to develop their potentials. With few authors includ-ing Native students their work, the findings and inferences of this meta-analysis Native youth are severely limited.
Conclusion
Achieving proportional representation for students identified as gifted is a goal worth attaining and an issue critical within the field of gifted education. Denying a large proportion of the public school children such services is not only socially unjust but also lacks foresight due to the population trends in the United States.
The results of this study provide evidence that disproportion-ality in gifted identification persists across all racial groups regardless of method used to identify the students as gifted. However, some areas in the United States have better propor-tional representation. And, verbal assessments have mitigated to some degree underrepresentation of Hispanic youth as gifted. Researchers should examine what the south and the southwest-ern regions of the United States are doing to close the gap in representation, and they should continue to use nontraditional means of identification together with other pathways to address disproportionality. The use of multiple identification methods and pathways would address the limitation of a single method. However, it is also important to adopt a more inclusive process in selecting gifted students for services. This can be done by providing services to students who are identified through any of the methods used by the schools and not requiring them to only be served when identified through multiple methods (McBee, Peters, & Waterman, 2014). Implementing talent development programs for traditionally underserved youth in which they can develop strengths, skills, and interests might help these students demonstrate and develop their potentials and thereby be identi-fied for further services. The representation gap will likely take years to close, but this should not discourage researchers and educators from seeking to improve proportionality in gifted programs.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, author-ship, and/or publication of this article.
References marked with an asterisk indicate studies included in the meta-analysis.Barnett, V., & Lewis, T. (1994). Outliers in statistical data (3rd
ed.). New York, NY: Wiley.*Bernal, E. M. (2002). Three ways to achieve a more equi-
table representation of culturally and linguistically differ-ent students in GT programs. Roeper Review, 24, 82-88. doi:10.1080/02783190209554134
Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2009). Introduction to meta-analysis. London, England: Wiley.
Borenstein, M., Hedges, L., Higgins, J., & Rothstein, H. (2014). Comprehensive meta-analysis (Version 3.1.1). Englewood, NJ: BIOSTAT.
Borland, J. H. (2003). The death of giftedness: Gifted education without gifted children. In J. H. Borland (Ed.), Rethinking gifted education (pp. 105-126). New York, NY: Teachers College Press.
*Bracken, B. A., & Brown, E. F. (2008). Early identification of high-ability students: Clinical assessment of behavior. Journal for the Education of the Gifted, 31, 403-426. doi:10.4219/jeg-2008-794
Briggs, C. J., Reis, S. M., & Sullivan, E. E. (2008). A national view of promising practices and programs for culturally, linguisti-cally, and ethnically diverse gifted and talented students. Gifted Child Quarterly, 52, 131-145. doi:10.1177/0016986208316037
Brown, A., & Lopez, M. H. (2013). Mapping the Latino population, by state, county, and city. Retrieved from http://www.pewhis-panic.org/2013/08/29/mapping-the-latino-population-by-state-county-and-city/
*Brulles, D., Peters, S. J., & Saunders, R. (2012). Schoolwide math-ematics achievement within the gifted cluster grouping model. Journal of Advanced Academics, 23, 200-216. doi:10.1177/1932202X12451439
Bush, G. W. (2001). No Child Left Behind. Washington, DC: Office of the President of the United States.
Callahan, C. M. (2005). Identifying gifted students from under-represented populations. Theory Into Practice, 44, 98-104. doi:10.1207/s15430421tip4402_4
Callahan, C. M., & Miller, E. M. (2005). A child-responsive model of giftedness. In R. J. Sternberg & J. E. Davidson (Eds.), Conceptions of giftedness (pp. 38-51). Cambridge, England: Cambridge University Press.
Card, N. A. (2012). Applied meta-analysis for social science research. New York, NY: Guilford.
*Carman, C. A., & Taylor, D. K. (2010). Socioeconomic status effects on using the Naglieri Nonverbal Ability Test (NNAT) to identify the gifted/talented. Gifted Child Quarterly, 54, 75-84. doi:10.1177/0016986209355976
*Civil Rights Data Collection. (2010). 2009 Civil rights data col-lection: Estimated values for United States (Office of Civil Rights, United States). Retrieved from http://ocrdata.ed.gov/statenationalestimations/projections_2009_10
Cooper, H. (2010). Research synthesis and meta-analysis: A step-by-step approach (4th ed.). Thousand Oaks, CA: Sage.
Cooper, H., Hedges, L. V., & Valentine, J. C. (Eds.). (2009). The handbook of research synthesis and meta-analysis. New York, NY: Russell Sage Foundation.
*Cribbs, D. K. (2009). Fourth grade teachers’ perceptions of gift-edness in Anglo and African American students (Doctoral dis-sertation). Available from ProQuest LLC. (3401320)
*Cross, T. L., Neumeister, K. L. S., & Cassady, J. C. (2007). Psychological types of academically gifted adolescents. Gifted Child Quarterly, 51, 285-294. doi:10.1177/0016986207302723
*Curby, T. W., Rudasill, K. M., Rimm-Kaufman, S. E., & Konold, T. R. (2008). The role of social competence in predicting gifted enrollment. Psychology in the Schools, 45, 729-744. doi:10.1002/pits.20338
Daniels, V. I. (1998). Minority students in gifted and special education programs: The case for educational equity. Journal of Special Education, 32, 41-43. doi:10.1177/002246699803200107
Donovan, M. S., & Cross, C. T. (Eds.). (2002). Minority students in special and gifted education. Washington, DC: National Academy Press.
Erwin, J. O., & Worrell, F. C. (2012). Assessment practices and the underrepresentation of minority students in gifted and talented education. Journal of Psychoeducational Assessment, 30, 74-87. doi:10.1177/0734282911428197
*Esquierdo, J. J., & Arreguín-Anderson, M. (2012). The “invis-ible” gifted and talented bilingual students: A current report on enrollment in GT programs. Journal for the Education of the Gifted, 35, 35-47. doi:0162353211432041.
Feldhusen, J. F., & Jarwan, F. A. (2000). Identification of gifted and talented youth for educational programs. In K. A. Heller, F. J. Mönks, R. J. Sternberg, & R. F. Subotnik (Eds.), International handbook of giftedness and talent (pp. 271-282). London, England: Elsevier.
Ford, D. (1998). The underrepresentation of minority students in gifted education: Problems and promises in recruit-ment and retention. Journal of Special Education, 32, 4-14. doi:10.1177/002246699803200102
*Ford, D. (2014). Segregation and the underrepresentation of blacks and Hispanics in gifted education: Social inequality and deficit paradigms. Roeper Review, 36, 143-154. doi:10.1080/02783193.2014.919563
Ford, D. Y., Grantham, T. C., & Whiting, G. W. (2008). Culturally and linguistically diverse students in gifted education: Recruitment and retention issues. Exceptional Children, 74, 289-306. doi:10.1177/001440290807400302
Ford, D. Y., & Harris, J. J. (1994). Promoting achievement among gifted Black students: The efficacy of new definitions and identification practices. Urban Education, 29, 202-229. doi:10.1177/0042085994029002006
Gallagher, J. J. (2004). No Child Left Behind and gifted education. Roeper Review, 26, 121-123. doi:10.1080/ 02783190409554255
Gentry, M. (2006). No Child Left Behind: Neglecting excellence. Roeper Review, 29, 24-27. doi:10.1080/02783190609554380
Gentry, M. (2009). Myth 11: A comprehensive continuum of gifted education and talent development services. Gifted Child Quarterly, 53, 262-265. doi:10.1177/0016986209346937
Gentry, M., Fugate, C. M., Wu, J., & Castellano, J. (2014). Gifted Native American students: Literature, lessons, and future directions. Gifted Child Quarterly, 58, 98-110. doi:10.1177/0016986214521660
Gentry, M., Peters, S., Pereira, N., McIntosh, J., & Fugate, C. (2015). HOPE Teacher Rating Scale (manual): Involving teachers in equitable identification of gifted and talented stu-dents in K-12. Waco, TX: Prufrock Press.
*Giessman, J. A., Gambrell, J. L., & Stebbins, M. S. (2013). Minority performance on the Naglieri Nonverbal Ability Test, versus the Cognitive Abilities Test, Form 6: One gifted program’s experience. Gifted Child Quarterly, 57, 101-109. doi:10.1177/0016986213477190
Grantham, T. C. (2003). Increasing Black student enrollment in gifted programs: An exploration of the Pulaski county special school district’s advocacy efforts. Gifted Child Quarterly, 47, 46-65. doi:10.1177/001698620304700106
Grantham, T. C. (2011). New directions for gifted Black males suf-fering from bystander effects: A call for upstanders. Roeper Review, 33, 263-273. doi:10.1080/02783193.2011.603114
Hedges, L. V., & Maier, K. S. (2013). Meta-analysis. In M. A. Scott, J. S. Simonoff, & B. D. Marx (Eds.), The Sage handbook of multi-level modeling (pp. 487-502). Thousand Oaks, CA: Sage.
Hedges, L. V., Tipton, E., & Johnson, M. C. (2010). Robust vari-ance estimation in meta-regression with dependent effect size estimates. Research Synthesis Methods, 1, 39-65. doi:10.1002/jrsm.5
Heller, K. A., & Hany, E. (2004). Identification of gifted and tal-ented students. Psychology Science, 46, 302-323.
Higgins, J., & Thompson, S. G. (2002). Quantifying heterogene-ity in a meta-analysis. Statistics in Medicine, 21, 1539-1558. doi:10.1002/sim.1186
Hopkins, A., & Garrett, K. (2010). Separate and unequal: The underrepresentation of African American students in gifted and talented programs. Black History Bulletin, 73, 24-30.
Hox, J. J., & Leeuw, E. D. d. (2003). Multilevel models for meta- analysis. In S. P. Reise & N. Duan (Eds.), Multivariate applica-tions: Multilevel modeling: Methodological advances, issues, and applications (pp. 90-111). Mahwah, NJ: Lawrence Erlbaum.
Huffcutt, A. I., & Arthur, W. (1995). Development of a new outlier statistic for meta-analytic data. Journal of Applied Psychology, 80, 327-334. doi:10.1037/0021-9010.80.2.327
*Illinois State Board of Education. (2003). Gifted education in Illinois school year 2003. Retrieved from http://www.isbe.net/research/pdfs/gifted_stats_02-03.pdf
*Jones, N. L. (2010). Culturally and linguistically diverse students and gifted and talented programs (Doctoral dissertation). Available from ProQuest LLC. (3428485)
*Kettler, T. (2014). Critical thinking skills among elementary school students comparing identified gifted and general educa-tion student performance. Gifted Child Quarterly, 58, 127-136. doi:10.1177/0016986214522508
Kettler, T., Russel, J., & Puryear, J. S. (2015). Inequitable access to gifted education: Variance in funding and staffing based on locate and contextual school variables. Journal for the Education of the Gifted, 38, 99-117. doi:10.1177/0162353215578277
Konstantopoulos, S., Modi, M., & Hedges, L. V. (2001). Who are America’s gifted? American Journal of Education, 109, 344-382. doi:10.1086/444275
Krisel, S. (2012). Characteristics of gifted children as a guide to identification. In S. L. Hunsaker (Ed.), Identification: The the-ory and practice of identifying students for gifted and talented education services (pp. 75-97). Mansfield Center, CT: Creative Learning Press.
Lakin, J. M., & Lohman, D. F. (2011). The predictive accu-racy of verbal, quantitative, and nonverbal reasoning tests: Consequences for talent identification and program diver-sity. Journal for the Education of the Gifted, 34, 595-623. doi:10.1177/016235321103400404
*Lange, R. W. (2004). Sources of underrepresentation in the iden-tification of African American students in elementary school gifted and talented education: A case study of one subur-ban school district’s gifted program (Doctoral dissertation). Available from UMI Microform. (3157995)
Lee v. Lee County Board of Education. 963 F. Supp. 1122 (M.D. Ala. 1997).
*Lee, S. Y., & Olszewski-Kubilius, P. (2006). Comparisons between talent search students qualifying via scores on stan-dardized tests and via parent nomination. Roeper Review, 28, 157-166. doi:10.1080/02783190609554355
*Lewis, J. D., DeCamp-Fritson, S. S., Ramage, J. C., McFarland, M. A., & Archwamety, T. (2007). Selecting for ethnically diverse children who may be gifted using Raven’s Standard Progressive Matrices and Naglieri Nonverbal Abilities Test. Multicultural Education, 15, 38-42.
*Lindsey, C. D. (2012). An investigation of the combined assess-ments used as entrance criteria for a gifted English middle school program (Doctoral dissertation). Available from ProQuest LLC. (3525542)
*Linn-Cohen, R., & Hertzog, N. B. (2007). Unlocking the GATE to differentiation: A qualitative study of two self-contained gifted classes. Journal for the Education of the Gifted, 31, 227-259. doi:10.4219/jeg-2007-677
Lohman, D. F. (2005). Review of Naglieri and Ford (2003): Does the Naglieri Nonverbal Ability Test identify equal proportions of high-scoring White, Black, and Hispanic students? Gifted Child Quarterly, 49, 19-28. doi:10.1177/001698620504900103
Lohman, D. F. (2011). Introducing CogAT Form 7. Retrieved from http://riversidepublishing.com/products/cogAT7/pdf/CogSpe_v59-28-11.pdf
*Lohman, D. F., & Gambrell, J. L. (2012). Using nonverbal tests to help identify academically talented children. Journal of Psychoeducational Assessment, 30, 25-44. doi:10.1177/0734 282911428194
Lohman, D. F., & Korb, K. A. (2006). Gifted today but not tomor-row? Longitudinal changes in ability and achievement during elementary school. Journal for the Education of the Gifted, 29, 451-484. doi:10.4219/jeg-2006-245
*Lohman, D. F., Korb, K. A., & Lakin, J. M. (2008). Identifying academically gifted English-language learners using nonverbal tests: A comparison of the Raven, NNAT, and CogAT. Gifted Child Quarterly, 52, 275-296. doi:10.1177/0016986208321808
Lohman, D. F., & Nicpon, M. F. (2012). Ability testing & talent identification. In S. L. Hunsaker (Ed.), Identification: The theory and practice of identifying students for gifted and tal-ented education services (pp. 287-335). Mansfield Center, CT: Creative Learning Press.
*Makel, M. C., Li, Y., Putallaz, M., & Wai, J. (2011). High-ability students’ time spent outside the classroom. Journal of Advanced Academics, 22, 720-749. doi:10.1177/1932202X11424880
*Matthews, M. S. (2006). Gifted students dropping out: Recent findings from a southeastern state. Roeper Review, 28, 216-223. doi:10.1080/02783190609554367
*McBee, M. T. (2006). A descriptive analysis of referral sources for gifted identification screening by race and socioeconomic sta-tus. Journal of Advanced Academics, 17, 103-111. doi:10.4219/jsge-2006-686
McBee, M. T., Peters, S. J., & Waterman, C. (2014). Combining scores in multiple-criteria assessment systems: The impact of combination rule. Gifted Child Quarterly, 58, 69-89. doi:10.1177/0016986213513794
McClain, M. C., & Pfeiffer, S. (2012). Identification of gifted stu-dents in the United States today: A look at state definitions, policies, and practices. Journal of Applied School Psychology, 28, 59-88. doi:10.1080/15377903.2012.643757
McKenzie, J. A. (1986). The influence of identification practices, race and SES on the identification of gifted students. Gifted Child Quarterly, 30, 93-95. doi:10.1177/001698628603000210
*Mills, C. J., & Tissot, S. L. (1995). Identifying academic poten-tial in students from under-represented populations: Is using the Ravens Progressive Matrices a good idea? Gifted Child Quarterly, 39, 209-217. doi:10.1177/001698629503900404
Milner, H. R., & Ford, D. Y. (2007). Cultural considerations in the underrepresentation of culturally diverse elementary students in gifted education. Roeper Review, 29, 166-173.
*Monrad, D. M., McGuiness, P. L., Chandler, D., Gallant, D. J., Gareau, S. J., & Rawls, A. M. (2005). A descriptive study of South Carolina’s gifted and talented program. Retrieved from http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=2510169ADC511540CA7778556BC24AC8?doi=10.1.1.693.2241&rep=rep1&type=pdf
*Naglieri, J. A., & Ford, D. Y. (2003). Addressing underrepresenta-tion of gifted minority children using the Naglieri Nonverbal Ability Test (NNAT). Gifted Child Quarterly, 47, 155-160. doi:10.1177/001698620304700206
Naglieri, J. A., & Ford, D. Y. (2005). Increasing minority chil-dren’s participation in gifted classes using the NNAT: A response to Lohman. Gifted Child Quarterly, 49, 29-36. doi:10.1177/001698620504900104
National Association for Gifted Children. (n.d.). Definitions of gift-edness. Retrieved from http://www.nagc.org/resources-publi cations/resources/definitions-giftedness
National Association for the Education of Young Children. (2009). Developmentally Appropriate Practice in Early Childhood Programs Serving Children from Birth through Age 8. Retrieved from: https://www.naeyc.org/sites/default/files/globally-shared/downloads/PDFs/resources/position- statements/PSDAP.pdf
National Association for Gifted Children. (2010). Redefining gift-edness for a new century: Shifting the paradigm. Retrieved from http://www.nagc.org/sites/default/files/Position%20State ment/Redefining%20Giftedness%20for%20a%20New%20Century.pdf
National Association for Gifted Children. (2011). Identifying and serving culturally and linguistically diverse gifted stu-dents. Retrieved from http://www.nagc.org/sites/default/files/Position%20Statement/Identifying%20and%20Serving%20Culturally%20and%20Linguistically.pdf
National Association for Gifted Children. (2013). State definitions of giftedness. Retrieved from http://www.nagc.org.442elmp01 .blackmesh.com/sites/default/files/Advocacy/State%20defini tions%20(8-1-13).pdf
National Association for Gifted Children & The Council of State Directors of Programs for the Gifted. (2015). 2014-2015 State of the states in gifted education: Policy and practice data. Retrieved from https://www.nagc.org/sites/default/files/key%20reports/2014-2015%20State%20of%20the%20States%20(final).pdf
*Neumeister, K. L. S., Adams, C. M., Pierce, R. L., Cassady, J. C., & Dixon, F. A. (2007). Fourth-grade teachers’ perceptions of giftedness: Implications for identifying and serving diverse gifted students. Journal for the Education of the Gifted, 30, 479-499. doi:10.4219/jeg-2007-503
*Office for Civil Rights. (2015). 2009-10 National and state esti-mations. Retrieved from http://ocrdata.ed.gov/StateNational Estimations/Projections_2009_10
Olkin, I., & Gleser, L. (2009). Stochastically dependent effect sizes. In H. Cooper, L. V. Hedges, & J. C. Valentine (Eds.), The handbook of research synthesis and meta-analysis (2nd ed., pp. 357-376). New York, NY: Russell Sage Foundation.
*Olszewski-Kubilius, P., Lee, S. Y., Ngoi, M., & Ngoi, D. (2004). Addressing the achievement gap between minority and nonminority children by increasing access to gifted pro-grams. Journal for the Education of the Gifted, 28, 127-158. doi:10.1177/016235320402800202
Olszewski-Kubilius, P., & Limburg-Weber, L. (1999). Options for middle school and secondary level gifted students. Journal of Advanced Academics, 11, 4-10. doi:10.4219/jsge-1999-615
*Olszewski-Kubilius, P., & Turner, D. (2002). Gender differ-ences among elementary school-aged gifted students in achievement, perceptions of ability, and subject prefer-ence. Journal for the Education of the Gifted, 25, 233-268. doi:10.1177/016235320202500303
Passow, A. H., & Frasier, M. M. (1994). Toward improving iden-tification of talent potential among minority and disadvantage students. Roeper Review, 18, 198-202. doi:10.1080/0278319 9609553734
*Peairs, K. F., Eichen, D., Putallaz, M., Costanzo, P. R., & Grimes, C. L. (2011). Academic giftedness and alcohol use in early adolescence. Gifted Child Quarterly, 55, 95-110. doi:10.1177/0016986210392220
*Pendarvis, E., & Wood, E. W. (2009). Eligibility of historically underrepresented students referred for gifted education in a rural school district: A case study. Journal for the Education of the Gifted, 32, 495-514. doi:10.1177/016235320903200403
Perleth, C., Schatz, T., & Mönks, F. J. (2000). Early identification of high ability. In K. A. Heller, F. J. Mönks, R. J. Sternberg, & R. F. Subotnik (Eds.), International handbook of giftedness and talent (2nd ed., pp. 297-316). Oxford, England: Pergamon.
*Peters, S. J., & Gentry, M. (2012). Group-specific norms and teacher-rating scales: Implications for underrepresentation. Journal of Advanced Academics, 23, 125-144. doi:10.1177/1932202X12438717
Peters, S. J., Matthews, M., McBee, M., & McCoach, D. B. (2013). Beyond gifted education: Designing and implementing advanced academic programs. Waco, TX: Prufrock Press.
Pfeiffer, S. I. (2012). Current perspectives on the identification and assessment of gifted students. Journal of Psychoeducational Assessment, 30, 3-9. doi:10.1177/0734282911428192
Pfeiffer, S. I., & Blei, S. (2008). Gifted identification beyond the IQ test: Rating scales and other assessment procedures. In
S. I. Pfeiffer (Ed.), Handbook of giftedness in children (pp. 177-198). New York, NY: Springer.
*Pfeiffer, S. I., & Jarosewich, T. (2007). The Gifted Rating Scales–School form: An analysis of the standardization sample based on age, gender, race, and diagnostic efficiency. Gifted Child Quarterly, 51, 39-50. doi:10.1177/0016986206296658
*Pfeiffer, S. I., Petscher, Y., & Jarosewich, T. (2007). The Gifted Rating Scales—Preschool/kindergarten form: An analysis of the standardization sample based on age, gender, and race. Roeper Review, 29, 206-211. doi:10.1080/02783190709554410
*Pierce, R. L., Adams, C. M., Speirs Neumeister, K. L., Cassady, J. C., Dixon, F. A., & Cross, T. L. (2006). Development of an identification procedure for a large urban school cor-poration: Identifying culturally diverse and academically gifted elementary students. Roeper Review, 29, 113-118. doi:10.1080/02783190709554394
*Pierce, R. L., Cassady, J. C., Adams, C. M., Speirs Neumeister, K. L., Dixon, F. A., & Cross, T. L. (2011). The effects of cluster-ing and curriculum on the development of gifted learners’ math achievement. Journal for the Education of the Gifted, 34, 569-594. doi:10.1177/016235321103400403
*Ratcliff, N. J., Jones, C. R., Costner, R. H., Knight, C., Disney, G., Savage-Davis, E., . . . Hunt, G. H. (2012). No need to wait for Superman: A case study of one unique high school. Journal for the Education of the Gifted, 35, 391-411. doi:10.1177/0162353212459256
Raudenbush, S. W., Fotiu, R. P., & Cheong, Y. F. (1998). Inequality of access of educational resources: A national report card for eighth-grade math. Educational Evaluation and Policy Analysis, 20, 253-267. doi:10.3102/01623737020004253
Raven, J. (2000). The Raven’s progressive matrices: Change and stability over culture and time. Cognitive Psychology, 41, 1-48. doi:10.1006/cogp.1999.0735
Reis, S. M., & McCoach, D. B. (2000). The underachievement of gifted students: What do we know and where do we go? Gifted Child Quarterly, 44, 152-170. doi:10.1177/001698620004400302
Renzulli, J. S. (1978). What makes giftedness? Reexamining a defi-nition. Phi Delta Kappan, 60, 180-184, 261.
Renzulli, J. S. (2005). The three-ring conception of giftedness. In R. J. Sternberg & J. E. Davidson (Eds.), Conceptions of gift-edness (pp. 246-279). Cambridge, England: Cambridge University Press.
Renzulli, J. S., & Reis, S. M. (1991). The reform movement and the quiet crisis in gifted education. Gifted Child Quarterly, 35, 26-35. doi:10.1177/001698629103500104
Renzulli, J. S., & Reis, S. M. (2012). Defensible and doable: A prac-tical multiple-criteria gifted program identification system. In S. L. Hunsaker (Ed.), Identification: The theory and practice of identifying students for gifted and talented education services (pp. 25-56). Mansfield Center, CT: Creative Learning Press.
*Robertson, J. C. (2013). Self-concept, school satisfaction, and other selected correlates of subjective well-Being for advanced high school learners enrolled in two challenging academic set-tings. Journal for the Education of the Gifted, 36, 461-486. doi:10.1177/0162353213506068
*Rowe, E. W., Miller, C., Ebenstein, L. A., & Thompson, D. F. (2012). Cognitive predictors of reading and math achievement among gifted referrals. School Psychology Quarterly, 27, 144-153. doi:10.1037/a0029941
*Sarouphim, K. M. (2002). DISCOVER in high school: Identifying gifted Hispanic and Native American students. Journal of Advanced Academics, 14, 30-38. doi:10.4219/jsge-2002-385
*Sarouphim, K. M. (2004). DISCOVER in middle school: Identifying gifted minority students. Journal of Advanced Academics, 15, 61-69. doi:10.4219/jsge-2004-446
*Scott, M. S., & Delgado, C. F. (2005). Identifying cognitively gifted minority students in preschool. Gifted Child Quarterly, 49, 199-210. doi:10.1177/001698620504900302
Shaklee, B. D. (1992). Identification of young gifted stu-dents. Journal for the Education of the Gifted, 15, 134-144. doi:10.1177/016235329201500203
*Shaunessy-Dedrick, E., Evans, L., Ferron, J., & Lindo, M. (2015). Effects of differentiated reading on elementary students’ read-ing comprehension and attitudes toward reading. Gifted Child Quarterly, 59, 91-107. doi:10.1177/0016986214568718
Steenbergen-Hu, S., & Olszewski-Kubilius, P. (2016). How to con-duct a good meta-analysis in gifted education. Gifted Child Quarterly, 60, 134-154. doi:10.1177/0016986216629545
Stephen, K. R., Dudley, J., & Karnes, F. A. (2012). Identification of the gifted: An overview of legal issues. In S. L. Hunsaker (Ed.), Identification: The theory and practice of identifying stu-dents for gifted and talented education services (pp. 101-119). Mansfield Center, CT: Creative Learning Press.
Sternberg, R. J., & Davidson, J. E. (2005). Conceptions of gifted-ness. Cambridge, England: Cambridge University Press.
Subotnik, R. F., Olszewski-Kubilius, P., & Worrell, F. C. (2011). Rethinking giftedness and gifted education: A pro-posed direction forward based on psychological science. Psychological Science in the Public Interest, 12, 3-54. doi:10.1177/1529100611418056
*Suldo, S. M., & Shaunessy-Dedrick, E. (2013). Changes in stress and psychological adjustment during the transition to high school among freshmen in an accelerated curriculum. Journal of Advanced Academics, 24, 195-218. doi:10.1177/1932202X13496090
Sullivan, A. L. (2011). Disproportionality in special education identi-fication and placement of English language learners. Exceptional Children, 77, 317-334. doi:10.1177/001440291107700304
Sutton, A. J., Abrams, K. R., Jones, D. R., Sheldon, T. A., & Song, F. (2000). Methods for meta-analysis in medical research. New York, NY: Wiley.
Swanson, J. D. (2007). Policy and practice: A case study of gifted education policy implementation. Journal for the Education of the Gifted, 31, 131-164. doi:10.4219/jeg-2007-679
*Swiatek, M. A., & Lupkowski-Shoplik, A. (2003). Elementary and middle school student participation in gifted programs: Are gifted students underserved? Gifted Child Quarterly, 47, 118-130. doi:10.1177/001698620304700203
*Terry, A. W. (2003). Effects of service learning on young, gifted adolescents and their community. Gifted Child Quarterly, 47, 295-308. doi:10.1177/001698620304700406
*Texas Education Agency. (2011). Advanced placement and International Baccalaureate examination results in Texas 2009-10. Retrieved from http://tea.texas.gov/acctres/ap_ib_index.html
*Texas Education Agency. (2015). Public Education Information Management System: School district data. Retrieved from tea.texas.gov/Reports_and_data/school_district_data/
Tukey, J. W. (1960). A survey of sampling from contaminated distributions. In I. Olkin, S. G. Ghurye, W. Hoeffding, W. G. Madow, & H. B. Mann (Eds.), Contributions to probability and statistics: Essays in honor of Harold Hotelling (pp. 448-485). Stanford, CA: Stanford University Press.
Tukey, J. W. (1977). Exploratory data analysis. Reading, MA: Addison-Wesley.
Tyler, R. W. (1949). Basic principles of curriculum and instruction. Chicago, IL: University of Chicago Press.
Van den Noortgate, W., López-López, J. A., Marín-Martínez, F., & Sánchez-Meca, J. (2013). Three-level meta-analysis of depen-dent effect sizes. Behavior Research Methods, 45, 576-594. doi:10.3758/s13428-012-0261-6
Van den Noortgate, W., López-López, J. A., Marín-Martínez, F., & Sánchez-Meca, J. (2015). Meta-analysis of multiple outcomes: A multilevel approach. Behavior Research Methods, 47, 1274-1294. doi:10.3758/s13428-014-0527-2
Van Tassel-Baska, J. (2005). Domain-specific giftedness: Applications in school and life. In R. J. Sternberg & J. E. Davidson (Eds.), Conceptions of giftedness (pp. 358-376). Cambridge, England: Cambridge University Press.
*Van Tassel-Baska, J., Feng, A. X., & De Brux, E. (2007). A study of identification and achievement profiles of performance task-identified gifted students over 6 years. Journal for the Education of the Gifted, 31, 7-34. doi:10.4219/jeg-2007-517
*Van Tassel-Baska, J., Feng, A. X., & Evans, B. L. (2007). Patterns of identification and performance among gifted students iden-tified through performance tasks a three-year analysis. Gifted Child Quarterly, 51, 218-231. doi:10.1177/0016986207302717
*Van Tassel-Baska, J., Johnson, D., & Avery, L. D. (2002). Using performance tasks in the identification of economi-cally disadvantaged and minority gifted learners: Findings from Project STAR. Gifted Child Quarterly, 46, 110-123. doi:10.1177/001698620204600204
*Warne, R. T., Anderson, B., & Johnson, A. O. (2013). The impact of race and ethnicity on the identification process for giftedness in Utah. Journal for the Education of the Gifted, 36, 487-508. doi:10.1177/0162353213506065
Watkins, M. W., Greenawalt, C. G., & Marcell, C. M. (2002). Factor structure of the Wechsler Intelligence Scale for Children–Third Edition among gifted students. Educational and Psychological Measurement, 62, 164-172.
*Wilson, H. E. (2015). Patterns of play behaviors and learning cen-ter choices between high ability and typical children. Journal of Advanced Academics, 26, 143-164. doi:10.1177/1932202X15577954
*Wilson, H. E., & Adelson, J. L. (2012). College choices of aca-demically talented secondary students. Journal of Advanced Academics, 23, 32-52. doi:10.1177/1932202X11430269
*Woitaszewski, S. A., & Aalsma, M. C. (2004). The contribution of emotional intelligence to the social and academic success of gifted adolescents as measured by the multifactor emotional intelligence scale-adolescent version. Roeper Review, 27, 25-30. doi:10.1080/02783190409554285
Wood, J. (2008). Methodology for dealing with duplicate study effects in a meta-analysis. Organizational Research Methods, 11, 79-95. doi:10.1177/1094428106296638
Yoon, S. Y., & Gentry, M. (2009). Racial and ethnic representation in gifted programs: Current status of and implications for gifted
174 Gifted Child Quarterly 62(2)
Asian American students. Gifted Child Quarterly, 53, 121-136. doi:10.1177/0016986208330564
Author Biographies
Jaret Hodges is a post doctoral associate at the Duke University Talent Identification Program. His research interests are in underrep-resented populations in gifted education and gifted education policy.
Juliana Tay is a doctoral candidate at Purdue University where she is pursuing a degree in gifted education. Her research interests include gifted art learners, issues in identification for giftedness, and evaluation of gifted programs.
Yukiko Maeda Ph.D. is an Associate Professor of Educational Psychology at Purdue University, specializing in research meth-odology. Her expertise includes research synthesis, meta-analy-sis and multilevel modeling. She has collaborated with educa-tional researchers in various disciplines as a data analyst and a
methodological expert to supervise quantitative data analyses and oversee the implementation of research design. She partici-pated in several research projects on gifted and talent develop-ment funded by U.S. DOE Javits Program and by NSF.
Marcia Gentry is the director of the Gifted Education Resource Institute and Professor of Educational Studies at Purdue University. Her research has focused on the use of cluster grouping and differ-entiation; the application of gifted education pedagogy to improve teaching and learning; student perceptions of school; and on non-traditional services and underserved populations. Marcia developed and studied the Total School Cluster Grouping Model and is engaged in continued research on its effects concerning student achievement and identification and on teacher practices. She is past chair of the AERA SIG, Research on Giftedness, Creativity, and Talent, actively participates in NAGC, frequently contributes to the gifted education literature, and regularly serves as a speaker and consultant.
