Can the Brigance K&1 Screen Detect Cognitive/Academic Giftedness When Used with Preschoolers from Economically Disadvantaged Backgrounds?

The need for early identification procedures to detect economically disadvantaged children who are gifted and academically talented has been repeatedly identified in the literature (i.e., Borland & Wright, 1994; Hadaway & Marek-Shroer, 1992). Although it is asserted that the potential for giftedness exists in all cultural and socioecomic groups (Borland & Wright, 1994), national data have indicated that students from low socioeconomic backgrounds are underrepresented in gifted education programs (United States Department of Education, 1991).

Over a decade ago, Ehrlich (1986) highlighted several issues that may impact the accuracy with which commonly used procedures detect young gilled children from non-mainstream backgrounds. Nominations from teachers and parents, a popular method for identifying giftedness (Renzulli, Reis, & Smith, 1981), may not be valid with young children living in poverty for several reasons. First, teachers have limited information on which to form judgments about a child's intellectual skills and abilities. Second, teacher judgments of disadvantaged children may be biased (Ehrlich, 1986; Hadaway & MarekSchoer, 1992). Third, teachers of nonmainstream children may find it difficult to think of them as gifted because the children, in addition to their academic talents, may exhibit social behaviors that are incompatible with traditional views of students with high cognitive competence. Finally, economically disadvantaged parents may mistrust the educational system and/or may not have sufficient knowledge and skills to successfully advocate for their young children (Woods & Achey, 1990).

Information obtained from standardized measures serves as another means for identifying gifted students. However, most of these measures require a highly trained examiner and their use with large numbers of children may be very costly. Futhermore, the appropriateness of these measures for non-mainstream children has been questioned by some (for a brief review see Hadaway & Marek-Schoer, 1992). Nevertheless, others (i.e., Ehrlich, 1986) have commented favorably on the usefulness of certain established tests of cognitive skills with children from disadvantaged backgrounds. Unfortunately, because of the high cost, routine screening of children with these measures (i.e., the WISC-R) is an unrealistic option for many programs.

Early screening tests form a group of instruments that, with few exceptions (e.g., Glascoe, 1996), have not been evaluated for their appropriateness in detecting early academic talent. This is not surprising, considering that a major purpose of these tests is the early detection of young children who may be at-risk for learning difficulties. Developmental screens are scanning devices that can be administered to young children in a relatively short time (i.e., 10-20 minutes, depending on the screen). They do not require highly trained examiners and are a preliminary step in the early identification process. A positive result is used to flag children who may be referred for more extensive assessment intended to provide accurate identification as well as appropriate placement and program planning.

Despite criticisms related to these tests narrow focus and psychometric properties (Lindsay & Wedell, 1982; Meisels, 1987; May & Kundert, 1997; Mercer, Algozzine, & Trifiletti, 1988; Shepard & Graue, 1993), early screens are widely used. Several studies have documented that early screening was mandated in several states and in several school districts in the remaining states (Cannella & Reif, 1989; Gnezda & Bolig, 1988; Shepard, Taylor, & Kagan, 1996). Although information about the prevalence of early screening measures in preschool programs is not available, early screening is a requirement set forth by Project Head Start, one of the nation's largest, federally supported early intervention programs for disadvantaged preschoolers. Head Start performance standards stipulate early screening of children within 45 days of their enrollment (Administration for Children and Families, 1999). Early screening data, therefore, may be among the first pieces of information collected by Head Start centers on children's skills across both cognitive and non-cognitive domains. Because information from early screening tests may be used to guide the early identification and assessment process, there is a compelling need for studies to carefully explore the properties of screening measures with the Head Start population. This study was conducted in response to this need and explores the predictive value of the Brigance K & 1 screen, a brief, 12-subtest battery, used by a Head Start Center housed in a midwestern school district.

Thus far, the focus of prior work on the Brigance K&1 screen has been on the early identification of educationally at-risk children (e.g., Campbell, Schellinger, & Beer, 1991; Ellwein, Walsh, Eads, & Miller, 1991; Mantzicopoulos, 1999a, 1999b; Shearer, 1986, Wenner, 1995). However, Glascoe (1996), noted that this measure may also be effective in the detection of children with cognitive/academic talents. Of particular interest was her finding that the screen "readily identified as gifted children from different ethnic and socioeconomic backgrounds" (p. 24). Glascoe's (1996) work set the stage for more research to empirically establish the accuracy of the Brigance in the detection of early cognitive/academic talent with different subpopulations of children.

This investigation is part of a series of studies intended to provide information on the K&1 form of the Brigance with different samples of low-income children (i.e., Mantzicopoulos, 1999a, 1999b). It was designed to explore differences between potentially gifted and non-gifted Head preschoolers on the screen and the level of accuracy with which the screen detects cognitive/academic giftedness. The sample came from a Head Start center that, unlike most Head Start centers, was part of a public school district. For this reason, the accuracy of the K&1 screen in the detection of potential cognitive/academic giftedness was examined in relationship to children's performance on outcome measures that were most meaningful with respect to their use within the school system in which this study was being conducted. These outcome measures included the Kaufman Assessment Battery for children, a standardized assessment of cognitive competence and the PPVT-R, a standardized achievement test. Although none of the children had been identified as gifted during the time of this investigation, performance on the K-ABC and the PPVT-R were among the indicators used by the public school administrators to make placement recommendations for gifted/talented programs for kindergarten and elementary school children.

The Brigance K&1 Screen: Overview and Prior Research

The Brigance K&1 screen is a 12-subtest instrument whose primary purpose is to identify young children in need of more comprehensive evaluation (Brigance, 1992). The test enjoys considerable popularity because of its seemingly multidimensional nature, the claim that it can be employed as a curriculum tool (Cohn, 1992), and its brevity and ease of administration (Shearer, 1986). This test consists of 12 subtests that measure: knowledge of personal data (5 items), color recognition (10 items), picture vocabulary (10 items), visual discrimination (10 items), visual-motor skills (5 items), gross-motor skills (10 items), rote counting from I to 10 (1 item), body parts identification (10 items), ability to follow verbal directions (2 items), numeral comprehension (5 items), ability to print first name (1 item), syntax and fluency (2 items). Each item is scored on a pass-fail basis. A raw score for each subtest is derived by summing across all items answered correctly by the child. The total score on each subtest is then calculated by multiplying the raw score with the weight assigned to that particular subtest.

Earlier reviews of the test highlighted the lack of normative data and cautioned school districts to undertake extensive validation procedures in order to avoid labeling, misdiagnoses, and further unnecessary testing (Cohn, 1992; Helfeldt, 1984). However, the screen's psychometric properties were evaluated in a recent technical report (Glascoe, 1997) which was based on a standardization sample of 408, 2- to 6-year old, ethnically diverse children. Twenty-two of the 408 children were Head Start participants.

The screen correlated moderately well with established measures of academic achievement (i.e., Woodcock-Johnson Psycho-Educational Battery-Revised) and of cognitive competence (i.e., Slosson Intelligence Test). The internal consistency reliability for the total scale was reported to be .99 for preschool and kindergarten children. Concurrent validity estimates, based on a sample of 75 children, indicated that the K&1 form of screen was 77% accurate in detecting children with disabilities and academic underachievement. This form was 75% accurate in identifying children without school-related difficulties. Moreover, of the 10 children who were classified as gifted/academically talented, 6 were correctly detected by the screen. Combined with teacher/parent ratings, the screen's accuracy in identifying gifted children increased to 80% (Glascoe, 1996).

Other research explored the extent to which performance on the screen varies across age, sex, and ethnic groups with somewhat mixed results. Two independent investigations concluded that boys and younger children scored consistently less well on the screen (Ellwein et al., 1991; Mantzicopoulos, 1999b). Additionally, Ellwein et al. (1991) also found that minority children scored less well than non-minority children at kindergarten. However, prior research with the Head Start sample used in this investigation identified neither sex nor age differences in performance on the K & 1 screen (Mantzicopoulos, 1999a).

Although exact data on the use of the Brigance screens with the Head Start population are not currently available, several sources suggest that the Brigance is frequently used across the nation (i.e., Gnezda & Bolig, 1988; May & Kundert, 1992; Meisels, Steele, & Quinn-Leering, 1993). Glascoe (1996) has more specifically noted that various forms of the Brigance screens are used with more than 700,000 children each year. Nevertheless, the extent to which the K & 1 form of the screen discriminates between possibly academically/cognitively gifted preschoolers from their non-gifted peers has not been clearly established. Glascoe's (1996) earlier findings with a diverse sample of two- to six-year-old children, convincingly substantiate the need for further research on this measure.

Method

Participants

Participants in the study were 134 children who attended a Head Start preschool program in a mid-western school district. These children represented 99% of the student population of the Head Start center during the school year, and were part of a longitudinal investigation of Head Start children's transition to public school. All children were screened with the kindergarten form of the Brigance K & 1 Screen during the second semester of Head Start (January-March). The age range of the sample at that time was from 52 to 69 months with a mean age of 61.6 months (SD = 3.37). There were 67 boys and 67 girls. Ethnicity data were available on 132 children, 102 (77.2%) of whom were Caucasian. Twenty-five (18.9%) children were African-American, and 5 (3.8%) were Other. All families noted that English was the language spoken in their homes and all met the federal poverty income guidelines necessary For entry into the Head Start program.

Although all 134 children were screened with the Brigance, data on the K-ABC were available on 132 children. Moroever, by the fall of kindergarten, 11 children had left the school district. Thus the PPVT-R assessments at the beginning of kindergarten, were based on a sample of 123 children. There were no differences between the children who left the study and those on whom assessments were made at kindergarten.

Instruments

Kaufman Assessment Battery for Children (K-ABC) The K-ABC (Kaufman & Kaufman, 1983a) is a carefully standardized measure that yields a Mental Processing Composite (MPC) that is comparable to the traditional intelligence quotient provided by other IQ tests. Correlations of the K-ABC MPC score with commonly used IQ tests (e.g., the Stanford-Binet and the Weschler Intelligence Scales) range from .36 to .72 for preschool and kindergarten samples (Kaufman & Kaufman, 1983b). The short form of the test, recommended for research purposes, was used in this study. This form is reported to correlate highly with the full battery. Validity coefficients, based on correlations between the subscale scores of the short form and the subscale scores derived from the full battery, range from .85 to .94 (Applegate & Kaufman, 1989 Kaufman & Applegate, 1988). The test provides standard scores with M = 100 and SD = 115. Studies have indicated that gifted children's average MPC scores are in the 120- to-130 range (Kaufman & Kaufman, 1983b).

Teachers' Ratings of Academic Competence Scale Developed by Harter and Pike (1981), this scale asks the teacher to rate each child on 6 items that communicate information about the child's competence in the cognitive/academic domain. Scored on a 4-point scale (1 = low; 4 = high), these items assess the child's knowledge of letters, counting skills, knowledge of colors, ability to solve puzzles, and overall school work. Because little has been published on this teacher rating scale, the alpha reliability coefficient was computed using the data from this sample. The resulting coefficient was .91.

Peabody Picture Vocabulary Test-Revised (PPVT-R) (Dunn & Dunn, 1981). The PPVT-R is an achievement test that measures a child's receptive vocabulary. It is considered to be a "reasonably good measure of scholastic aptitude" (Dunn & Dunn, 1981, p. 2) and may be used to identify bright children whose language at home is Standard English. Reliability studies report coefficients that range from .73 to .84 (Dunn & Dunn, 1981). Correlations with a number of established vocabulary, achievement, and IQ tests have been used to substantiate this test's construct validity. Median correlation coefficients range from .38 to .86. The test provides standard scores with M = 100 and SD = 115.

Procedure and Classification of High- and Average Ability Children

The Brigance K & 1 screen and the K-ABC were individually administered by research assistants in the winter of Head Start. In the fall of kindergarten, children were individually tested by school personnel with the PPVT-R, a test used by the school system to make readiness decisions and to guide the decision-making process for placing children in a gifted program. The PPVTR examiners were not aware of the children's prior performance on any measures, including the Brigance and the K-ABC.

At Head Start, children were classified into one of two groups on the basis of their scores on the Mental Processing Composite (MPC) of the K-ABC. A cutoff score of 115 or higher, identified 13 children who formed the group with possible cognitive/academic giftedness (CAG group). Their MPC scores ranged from 115 to 129 (M = 119.53; SD = 3.93). The MPC scores of the remaining children, who formed the nongifted (NG) group, ranged from 75.7 to 114 (M = 96.27, SD = 8.52). Gender, age, and ethnic group comparisons were conducted to explore differences between CAG and NG children. The groups were comparable on age and ethnicity. However, significantly more males (10/13) were identified in the CAG group than females (3/13) ([chi square] = 4.18, p = .04).

At the beginning of kindergarten, children were also classified into a CAG or NG group on the basis of their performance on the PPVT-R. Of the 124 children remaining in the study, 7 formed the CAG group. Their PPVT-R scores ranged from 115 to 130 (M= 119; SD = 2.93). The PPVT-R scores of the remaining 116 children, who formed the nongifted (NG) group, ranged from 90 to 114 (M = 96.27, SD = 12.76). The groups were comparable on age, sex, and ethnicity.

Results

Analyses

Two sets of analyses were conducted to examine the extent to which children's performance on the K &l screen's 12 subtests differed between the two groups and the concurrent and predictive accuracy of the screen in detecting early cognitive/academic giftedness. The first set of analyses included multivariate analysis of variance (MANOVA) using the SPSS GLM procedure (Norusis, 1993). Scores on the Brigance K & 1 screen's 12 subtests served as the dependent measures. Effect sizes were also computed and served as an index of the magnitude of differences between the groups. Effect sizes were derived by dividing the mean difference between the groups by the standard deviation of the NG group. This provided a conservative estimate because, as expected, the standard deviation of the NG group was consistently larger than the standard deviation of CAG group. Effect sizes were considered educationally significant when they were greater than +/- .20 standard deviation units (Cohen, 1988).

The second set of analyses were intended to communicate information on the level of accuracy with which the screen detects young Head Start children who may be cognitively/ academically gifted. As suggested by the literature on the evaluation of screening measures (i.e., Barnes, 1982; Gredler, 1997; Lichtenstein & Ireton, 1981; Meisels, 1989; Satz & Fletcher, 1988), the efficacy of the screen was examined through a series of classification analyses. These analyses evaluated the accuracy of the screen against outcomes obtained concurrently (i.e., performance on the K-ABC) and at a later point in children's schooling (i.e., performance on the PPVT-R at the beginning of kindergarten).

To prepare the data for the classification analyses, performance on the K & 1 screen was scored dichotomously (i.e., 1 = performance in the nongifted range; 2 = potential giftedness) using: the cutoff score for cognitive/academic giftedness provided in the Brigance technical manual (i.e., performance above a raw score of 93) (Glascoe, 1997); and a series of locally derived cutoff scores based on data from the current sample.

To compare the accuracy of the Brigance to the accuracy with which teachers' ratings detected potentially gifted children, teachers' ratings were also used as a predictor. To explore the usefulness of this predictor measure as thoroughly as possible, a series of cutoff scores in the teacher rating scale as predictors of possible giftedness were examined.

Following each analysis, the sensitivity, specificity, and overall hit rate associated with each classification table were evaluated. The sensitivity of the test served as an indicator of the test's accuracy in identifying those children who were predicted as potentially gifted by the screen and were later identified as CAG by the criterion measures (i.e., the K-ABC and the PPVT-R). Conversely, the specificity of the test provided information on the proportion of children who were predicted as potentially nongifted by the screen and were later identified as NG by the criterion measures. Finally, the hit rate of the test indicated the proportion of correctly classified children in the CAG and NG groups.

It is important to note that while sensitivity and specificity are indices of correct identifications, they are also related to specific errors in prediction. For example, a test with high sensitivity will result in few undereferrals (i.e., children who tested as potentially nongifted on the screen but who performed in the gifted range on a later criterion). For this reason, both accuracy indices and referral errors must be considered together to accurately evaluate the accuracy of a screening measure. Moreover, it has been suggested that to be of value, accuracy indices must approach (Carran & Scott, 1992) or exceed (Meisels, 1989) 80%.

In addition to the accuracy indices presented above, the phi ([Psi]) coefficients associated with each classification table were examine to gain information on the strength of the relationship between the Brigance prediction classification and the classification based on each performance variable.

Performance of Potentially Gifted and Nongifted Children on the Brigance

The MANOVA that examined differences between the CAG and NG group across the Brigance subtests yielded a statistically significant multivariate effect (F = 1.96 p =.03). CAG children performed significantly better than NG children on the Brigance as indicated by the data displayed in Table 1.

Table 1

Means, Standard Deviations, and Effect Sizes for the CAG and NG Group on the K & 1 Screen's Subtests

                              CAG             NG
                           (n = 13)        (n = 19)

                                                          ES
                           M      SD       M      SD

Personal Data            7.69    1.38    6.79    2.18     .41
Color Recognition        9.62     .65    8.18    2.11     .68
Picture Vocabulary      10.00     .00    9.76     .55     .44
Visual Discrimination    8.54    2.03    7.13    3.10     .45
Visual-Motor Skills      8.31    1.60    5.76    2.27    1.12
Gross-Motor Skills       9.00    1.00    7.88    2.16     .52
Rote Counting            5.00     .00    4.46    1.07     .50
Identify Body Parts      3.38     .89    3.11     .98     .28
Follows Directions       5.00     .00    4.90     .50     .20
Numeral Comprehension    9.08    2.25    6.67    3.86     .62
Prints Personal Data     3.46    2.40    1.22    2.16    1.04
Syntax and Fluency      10.00     .00    8.99    2.65     .38
Total Scale             89.08    4.22   74.86   13.12    1.08

Effect sizes were moderately large to large on the following four subtests: Prints Personal Data (ES = 1.04), Color Recognition (ES = .68), Visual-Motor Skills (ES = 1.12), and Numeral Comprehension (ES = .62). Moderate effect sizes were obtained on five subscales (Personal Data, Picture Vocabulary, Visual Discrimination, Gross-Motor Skills, and Identifies Body Parts). Finally, small yet educationally meaningful effect sizes were obtained on three subscales (Identifies Body Parts, Follows Directions, and Syntax/Fluency). It should be noted that on four of the twelve subscales (i.e., Picture Vocabulary, Rote Counting, Follows Directions, and Syntax/Fluency) there was no variance among children in the CAG group: All thirteen children obtained the highest possible score on each of these subscales.

Children's performance was also examined on the total, 71-item-scale. The difference between the group means was large. CAG children (M = 89.08, SD = 4.22) outperformed NG children (M = 74.86, SD - 13.12) on the screen by 1.08 standard deviations.

CAG children also differed from their NG peers on teacher ratings of cognitive competence (t = 2.07, p = .04). Teachers rated CAG children (M = 3.39, SD = .60) as significantly more competent in the cognitive/academic area than NG children (M = 2.96, SD = .72). However the effect size related to this group comparison was moderate (ES = .60).

Accuracy of the Brigance in the Early Detection of Cognitive/Academic Giftedness

Accuracy Related to Concurrent Performance on the K-ABC This analysis evaluated the accuracy of the K & 1 screen against performance on the KABC Cognitive Ability measure, administered concurrently with the screen during Head Start. Results are reported in Table 2.

Table 2

Classification Results Based on the K&I Screen Prediction and the K-ABC Outcome at Head Start

                                             K-ABC Classification
Brigance Cutoff Score   Prediction           NG               CAG

K&I Recommended         Nongifted            111                9

Cutoff                  Potentially Gifted     8                4

Local Cutoffs for Potentially Gifted Prediction

> [P.sub.60] / 81.5     Nongifted             78                1

                        Potentially Gifted    41               12

> [P.sub.65] / 84.0     Nongifted             84                1

                        Potentially Gifted    35               12

> [P.sub.70] / 85.0     Nongifted             88                2
                        Potentially Gifted    31               11

> [P.sub.75] / 86.5     Nongifted             95                4

                        Potentially Gifted    24                9

> [P.sub.80] / 88.0     Nongifted            100                4

                        Potentially Gifted    19                9
> [P.sub.85] / 89.0     Nongifted            105                5
                        Potentially Gifted    14                8

> [P.sub.90] / 91.0     Nongifted            110                9
                        Potentially Gifted     9                4

Prediction              Sens.   Spec.   Hit    [Psi]

Nongifted
                          .31     .93   .87   .25(*)
Potentially Gifted

Nongifted
                          .92     .66   .68   .35(**)
Potentially Gifted

Nongifted
                          .92     .71   .72   .39(**)
Potentially Gifted

Nongifted
                          .85     .74   .75   .38(**)
Potentially Gifted

Nongifted
                          .69     .80   .79   .33(**)
Potentially Gifted

Nongifted
                          .69     .84   .89   .39(**)
Potentially Gifted

Nongifted
                          .62     .88   .86   .40(**)
Potentially Gifted

Nongifted
                          .31     .92   .86   .23(*)
Potentially Gifted

Note. Sens. = Sensitivity; Spec. = Specificity

(*) p < .01; (**) p< .0001

The classification matrix based on the Brigance-recommended cutoff score of 93 or higher for possible giftedness/academic talent, was examined first. This classification resulted in low sensitivity (31%), high specificity (93%) and a high hit rate (87%). Use of this cutoff for making referrals for further assessment to identify possible giftedness/talent, would underrefer as notpotentially-gifted 9 of the 13 CAG children (69%)

Next, the classification accuracy of the screen was examined across a range of locally-derived prediction cutoffs for identifying possible giftedness/academic talent. The highest sensitivity level (i.e., correct identification of performance in the gifted range) was obtained when the prediction for potential giftedness was set at/above a raw score of 84 ([P.sub.65]). With this predictive criterion the test was 92% sensitive in identifying cognitively/academically gifted children. Also, the test was 71% specific in identifying those children with average-to-low scores on the performance outcome. It should be pointed out that the screen overreferred as potentially gifted 35 children (26.5% of the total sample) who performed in the nongifted range on the K-ABC. Finally, it is noteworthy that the screen underreferred as not-potentially-gifted only 1 of the 13 CAG children.

Inspection of the data shown in Table 2, further indicated that selection of more stringent prediction cutoffs (e.g., raw scores at/above 86.5) yielded less-than-optimal accuracy data. That is, while the specificity levels increased there was an accompanying increase in underreferral errors and a decrease in sensitivity levels.

The classification analyses using teachers' ratings of academic competence as the prediction criterion for possible cognitive/academic giftedness was also conducted across a range of cutoff scores. However, the highest indices obtained (sensitivity = 62% and specificity = 64%) were less-than-optimal and were associated with high misclassification rates.

Accuracy Related to Performance on the PPVT-R at Kindergarten The classification analyses based on the PPVT-R outcome are presented in Table 3. As noted earlier, these analyses are based on a sample of 123 children (7 CAG and 116 NG).

Table 3

Classification Results Based on the K&I Screen Prediction and the PPVT-R Outcome at the Beginning of Kindergarten

                                             PPVT-R Classification

Brigance Cutoff Score   Prediction           NG                CAG

K&I Recommended         Nongifted            108                 6

Cutoff                  Potentially Gifted     8                 1

Local Cutoffs for Potentially Gifted Prediction

> [P.sub.60] / 81.5     Nongifted             71                 1

                        Potentially Gifted    45                 6

> [P.sub.65] / 84.0     Nongifted             77                 1

                        Potentially Gifted    39                 6

> [P.sub.70] / 85.0     Nongifted             81                 1

                        Potentially Gifted    35                 6

> [P.sub.75] / 86.5     Nongifted             89                 1

                        Potentially Gifted    27                 6

> [P.sub.80] / 88.0     Nongifted             94                 1

                        Potentially Gifted    22                 6

> [P.sub.85] / 89.0     Nongifted             99                 2

                        Potentially Gifted    17                 5

> [P.sub.90] / 91.0     Nongifted            105                 5

                        Potentially Gifted    11                 2

Prediction              Sens.   Spec.   Hit     [Psi]

Nongifted
                          .14     .93    .89   .07
Potentially Gifted

Nongifted
                          .86     .61    .63   .22(*)
Potentially Gifted

Nongifted
                          .86     .66    .67   .25(**)
Potentially Gifted

Nongifted
                          .86     .70    .70   .27(**)
Potentially Gifted

Nongifted
                          .86     .78    .77   .33(***)
Potentially Gifted

Nongifted
                          .86     .81    .81   .37(***)
Potentially Gifted

Nongifted
                          .71     .85    .84   .34(**)
Potentially Gifted

Nongifted
                          .29     .91    .87   .14(*)
Potentially Gifted

Note. Sens.= Sensitivity; Spec. = Specificity

(*) p < .05; (**) p < .01; (***) p < .0001

Performance at/above a cutoff score of 88 ([P.sub.80) on the K & 1 screen, effectively identified those children with scores in the gifted range on the PPVT-R (sensitivity = 86%). In addition, this cutoff correctly identified 81% of the NG children (sensitivity = 81%). The overall hit rate associated with this classification was 81% ([Psi] = .37). Although use of this cutoff score resulted in 22 children being overreferred as potentially gifted (18% of the total sample), only one of the seven CAG children was underreferred by the screen as not-potentially-gifted.

As shown in Table 3, selection of more liberal prediction criteria for the identification of potential giftedness did not substantially alter the sensitivity of the screen but compromised its specificity and resulted in higher over-referral errors. For example, use of a raw score of 81.5 overreferred as potentially gifted 45 children (37% of the total sample). Conversely, use of more conservative prediction criteria (e.g. a cutoff score of 91 and the Brigance-recommended cutoff score of 93) produced high specificities with an accompanying decrease in the sensitivity levels. Such low sensitivities and high underreferral errors would counterindicate use of conservative cutoffs the early detection of giftedness among low-income children.

Use of Head Start teachers' ratings as predictors of PPVT-R performance at kindergarten yielded low sensitivities across a range of prediction cutoffs. The highest sensitivity level (57%) was associated with a high overreferral error (59% of the total sample) and a low specificity (37%).

Discussion

The results of this study provided support for the use of the Brigance K & 1 screen in the early identification of economically disadvantaged children with possible cognitive/academic giftedness. Potentially gifted Head Start children in this sample performed significantly better on the K & 1 screen than their average-ability peers. Educationally meaningful group differences were observed on all 12 subscales of the screen. Moreover, the group differences on the K & 1 total score were large and exceeded one standard deviation. Classification analyses further substantiated that performance on the Brigance accurately predicted a child's cognitive ability level as measured by the K-ABC. In addition, the predictive accuracy of the screen in relationship to the PPVT-R met the standards for screening measures set forth in the literature (i.e., Carran & Scott, 1992; Meisels, 1989).

It should be noted that these accuracy indices were obtained with cutoff-scores lower - than-recommended in the technical report for the screen (Glascoe, 1997). However, the technical report did not include data on the predictive value of the screen with potentially gifted children from low-income backgrounds. Rather, the recommended cut off score of 93 was derived from a diverse sample of 75 children, 10 of whom were considered to be cognitively/academically gifted. While Glascoe (1997) did comment that adjustments to the recommended cutoff scores may be needed when screening children who come from multiple risk backgrounds, there are no data to guide test users. Clearly, the results of this investigation substantiate the need to explore the levels at which the screening results may be used to increase the accuracy of referrals for further assessments intended to identify young children with giftedness.

The analysis of the teacher ratings did not support their use in the early detection of academic talent over the K & 1 screen. This finding parallels prior work in which it has been suggested that teacher reports may be unreliable unless teachers are given specific training on the characteristics of gifted students (Hadaway & Schroer, 1992). Possibly, the lower accuracy of the teacher ratings in this study may have been a function of the scale that the teachers were asked to complete. In earlier studies (e.g., Borland & Wright, 1994; Glascoe, 1996), teachers provided nominations for students they thought might be gifted, while in this study teachers simply rated children's cognitive competence. It was assumed that academically gifted children would be given the highest ratings by their teachers on a cognitive/academic competence measure. It is not clear, therefore, whether or not teachers would have been more accurate if they had been explicitly asked to nominate specific students.

A limitation of the study is the use of the K-ABC and the PPVT-R measure to determine early cognitive/academic giftedness. These measures present a narrow view of competence and may be considered a screening tool, much like the Brigance. However, that both these tests were used for decision-making in the public school school system, served as the rationale for their inclusion in the current investigation.

Another drawback of the study may be the use of liberal performance criteria (i.e., MPC scores of 115 or higher) in the classification of potentially gifted and nongifted children. It could be argued that these criterion cutoffs allowed for the inclusion of children who had potentially good cognitive and academic skills but who were not necessarily gifted. In response, the I 15- point cutoff score was considered appropriate for this population because economically disadvantaged children generally, tend to perform less well on standardized tests. Ideally, it would be important to document the predictive accuracy of the screen against a criterion of giftedness based on the number of children actually identified by public school or program procedures. Such procedures were not undertaken at the Head Start center, where a major concern was the identification and provision of appropriate services to those at-risk for school-related problems. Later, as children from this sample transitioned to public school only one of the 13 children in the potentially gifted subgroup (and none from the nongifted subgroup) was identified as gifted. This is not surprising, in light of the documented underidentification rates for economically disadvantaged and minority children in the public schools (United States Department of Education, 1991). Evidence on measures and procedures that facilitate the early detection of academic talent is one of the steps needed to improve this process.

In conclusion, there is evidence that when used with preschool children from disadvantaged backgrounds, the Brigance K & 1 screen effectively identifies those children who may be cognitively/academically gifted. The findings are preliminary, and considering the popularity of the screen, more empirical data are necessary to establish the efficacy of this measure in the early detection of academic talent.

REFERENCES

Administration for Children and Families (1999). Head Start program performance standards, [on line paper] Washington, DC: U.S. Department of Health and Human Services, Head Start Bureau. Available Internet Address: http//www2.acf.dhhs.gov/programs/hsb

Applegate, B., & Kaufman, A. S. (1989). Short form estimation of the K-ABC sequential and simultaneous processing for research and screening. Journal of Clinical Child Psychology, 18(4), 305-313.

Barnes, K. E. (1982). Preschool screening: The measurement and predict, on of children at-risk. Springfiled, IL: Charles C. Thomas.

Borland, J. H., & Wright, L. (1994). Identifying young, potentially gifted, economically disadvantaged students. Gifted Child Quarterly, 38(4), 164-171.

Brigance, A. (1992). Brigance K & 1 Screen (3rd ed.) North Billerica, MA: Curriculum Associates.

Brigance, A. (1991). Brigance K & 1 Screen-Revised: Technical report. North Billerica, MA: Curriculum Associates.

Campbell, E., Schelliager, T., & Beer, J. (1991). Relationships among the ready or not parental checklist for school readiness, the Brigance kindergarten and first grade screen, ant SRA scores. Perceptual and Motor Skills, 73, 859-862.

Cannella, G. S., & Reif, J. C. (1989). Mandating early childhood entrance/retention assessment: Practices in the 50 slates. Paper presented at the annual meeting of the American Educational Research Association, San Francisco.

Carran, D. T., & Scott, K. G. (1992). Risk assessment in preschool children: Research implications for the early identification of educational handicaps. Topics in Early Childhood Special Education. 12(2), 196-211.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (Rev. ed). New York: Academic Press. Cohn, M. E. (1992). Screening measures. In E. V. Nutall, I. Romero, & J. Kalesnik (Eds.), Assessing and screening preschoolers (pp. 83-98). Needham heights, MA: Allyn & Bacon.

Dunn, L. M.. & Dunn, L. M. (1981). Peabody Picture Vocabulary Test-Revised. Circle Pines, MN: American Guidance Service.

Ellwein, M. C,, Walsh, D. J., Eads II, G. M., & Miller, A. (1991). Using readiness tests to route kindergarten students: The snarled intersection of psychometrics. policy, and practice. Educational Evaluation and Policy Analysis, 13(2), 159-175.

Ehrlich, V. Z. (1986) Recognizing superior cognitive abilities in disadvantaged, minority, and other diverse populations, Journal of Children in Contemporary Society. 18, 55-70.

Glascoe, F. P. (1996). Can the Brigance screens detect children who are gifted and academically talented? Roeper Review, 9(1), 20-24.

Glascoe, F. P. (1997). Technical report for the Brigance screens. North Billerica, MA: Curriculum Associates.

Gnezda, M. T., &Bolig, R. (1988). A national survey of public school testing of pre-kindergarten and kindergarten children. Washington, DC: National Forum on the Future of Children and Families, National Research Council.

Gredler, G. R. (1997). Issues in early childhood screening and assessment. Psychology in the Schools, 34(2), 99-106.

Hadaway, N., & Marek-Schroer, M. F. (1992). Multidimensional assessment of the gifted minority student. Roeper Review. 15(2), 73-77.

Hatter, S., & Pike, R (1984). The pictorial scale of perceived competence and social acceptance for young children. Child Development, 55, 1969-1982.

Helfeldt, J. P. (1984)The Brigance K & 1 screen for kindergarten and first grade. The Reading Teacher, 37(9), 820-824.

Kaufman, A. S., & Applegate, B. (1988). Short forms of the K-ABC Mental Processing and achievement scales at ages 4 to 121/1 years for clinical and screening purposes. Journal o" Clinical Child Psychology, 17(4), 359-369.

Kaufman, A. S., & Kaufman, N. L. (1983a). Kaufman Assessment Battery for Children. Circle Pines, MN: American Guidance Service.

Kaufman, A. S., & Kaufman, N. L. (1983b). Kaufman Assessment Battery for Children: Interpretive manual. Circle Pines, MN: American Guidance Service.

Lichtenstein, R., & Ireton, H. (1991). Preschool screening for developmental and educational problems. In B. A. Bracken (Ed.), The psychoeducational assessment of preschool children (2nd ed., pp. 486-513). Needham Heights, MA: Allyn and Bacon.

Lindsay, G. A., & Wedell, K. (1982). The early identification of educationally "at-risk" children revisited. Journal of Learning Disabilities, 15, 212-217.

Mantzicopoulos, P. (1999a). Reliability and validity estimates of the Brigance K & 1 screen based on a sample of disadvantaged preschoolers. Psychology in the Schools, 36(1), 1-9.

Mantzicopoulos, P. (1999, April). Screening Head Start Children: Relationships of the Brigance K & 1 Screen to Early School Achievement and Special Education Placement. Paper presented at the annual meeting of the American Educational Research Association, Montreal.

Mantzicopoulos, P. Y. (1999b). Early screening of Head Start children: Relationships of the Brigance K & 1 Screen to Early School Achievement and Special Education Placement. Manuscript submitted for publication.

May, D.C., & Kundert, D. K. (1992). Kindergarten screening in New York State: Tests, purposes, and recommendations. Psychology in the Schools, 29, 35-41.

May, D.C., & Kundert, D. K. (1997). School readiness practices and children at risk: Examining the issues. Psychology in the Schools, 34(2), 73-84.

Meisels, S. (J 987). Uses and abuses of developmental screening and school readiness testing. Young Children. 42, 4-7.

Meisels, S. J. (1989) Can developmental screening tests identify children who are developmentally at-risk? Pediatrics, 83(4), 578-585.

Meisels, S. J., Steele, D. M., & Quinn-Leering, K 91993). Testing, tracking, and retaining young children: An analysis of research and social policy. In B. Spodek (Ed.), Handbook of research on the education of young children (pp. 279-292). New York: MacMillan.

Mercer, C. D., Algozzine, B., & Trifiletti, J. (1988). Early identification--An analysis of research. Learning Disability Quarterly, 2, 12-24.

Norusis, M. J. (1993). SPSS for Windows. Chicago, IL: SPSS Inc.

Renzulli, J. S., Reis, S. M., & Smith, L. H. (1981). The revolving door identification model. Mansfield, CT: Creative Learning Press.

Satz, P., & Fletcher, J. M. (1988). Early identification of learning disabled children: an olde problem revisited. Journal of Consulting and Clinical Psychology, 6, 824-829.

Schearer, M. (1986). A review of Chapter 53 screening. Brooklyn, NY: New York City Board of Education, Office of Educational Assessment (ERIC Document Reproduction Service, No. ED 287 236).

Shepard, L. A., & Graue, M. E. (1993). The morass of school readiness screening: Research on test use and test validity. In B. Spodek (Ed.), Handbook of research on the education of young children (pp, 293-305). New York: MacMillan.

Shepard, L. A., Taylor, G. A., & Kagan, S. L. (1996, October). Trends in early childhood assessment policies [on line paper]. Washington DC: National Education Goals Panel. Available Internet Address: http://www.negp.gov

United States Department of Education (1991). National educational longitudinal study 88. Final report: Gifted and talented education programs for eighth-grade public school students. Washington DC: Office of Planning, Budget, and Evaluation.

Wenner. G. (1995). Kindergarten screens as tools for the early identification of children at risk for remediation or grade retention. Psychology in the Schools, 31, 249-254.

Woods, S. B., & Achey, V. H. (1990). Successful identification of gifted racial/ethnic group students without changing classification requirements. Roeper Review. 13, 10-15.

Manuscript submitted July, 1998.

Revision accepted April, 1999.

Panayotan Mantzicopoulos is associate professor of Educational Psychology,Department of Educational Studies, Purdue University.