Gifted and highly gifted students: how do they score on the SB5?

The year 2003 ushered in revisions of the two intelligence scales most commonly used in determinations of giftedness--the Stanford-Binet (Roid, 2003) and the Wechsler Scales (Wechsler, 2003). Both scales reflect changes that go beyond simple item updates and new norms; there are refinements in the conceptualization of intelligence and corresponding changes in item subject matter. Because of these changes, use of these new tests may influence the number and type of students that are determined to be eligible for gifted programs.

In general, both tests have shifted toward use of the CHC model of intelligence (Carroll, 1993). Consistent with this emphasis, content on both tests includes items to measure both crystallized and fluid abilities. Because children identified as gifted show greater variability and lower overall performance on processing speed items such as those included on the WISC-IV (Wechsler, 1991) and on working memory items which are included in both tests (Roid, 2003), use of full-scale scores that place an increasing emphasis on these factors will likely exclude some children who would have been identified as gifted on earlier tests. The inclusion of items that are more dependent upon visual-spatial and nonverbal skills will increase scores upwards for children with these strengths while lowering scores of children who have strengths in the verbal-abstract-reasoning area more traditionally emphasized by those who provide programs for the gifted. Publishers of both scales have provided experimental index scores purported to be useful in emphasizing the reasoning aspects of intelligence. At this point, however, these scores have not been normed and validated on independent samples of gifted children.

The norming process itself may have an effect on intelligence scores for children in the gifted range. While renorming to counteract the Flynn effect (Flynn, 1984) may seem like an innocuous process, Kanaya, Scullin, and Ceci (2003), in their discussion of mental retardation diagnoses, maintain that the introduction of a newly normed IQ test can create havoc for a few years due to differences in IQ scores, which, in turn, influence determinations of program entry. Flynn (2000) documented the shifting percentage of people diagnosed as mentally retarded over the last 50 years from a high of 1 in 23 to a low of 1 in 213. These results appear to be due primarily to both an increase in IQ over time and variation in samples used to provide norms. A similar process may operate in determinations of giftedness. While it is unclear to what extent the Flynn effect applies to gifted populations (Flynn described scores in the gifted range as "eccentric" and excluded them from analysis in his 1984 article), it is certainly possible that a child designated as gifted on the basis of IQ the year before introduction of a new edition of an IQ test would no longer meet the criteria for placement if given the new edition. This can lead to the odd situation in which children who are assessed on a newer version of a test are excluded from program entry, while those assessed just before introduction of the new version gain admittance.

It is difficult to accurately norm scores on the end of the continuum. Because so few people have extremely high scores, there is a dependence upon statistical modeling for distributions of intelligence scores in the high ranges. A variety of researchers have questioned the accuracy of such models since the use of the first intelligence test in this country (Silverman, 1989; Terman, 1925; Wechsler, 1944) and have suggested that theoretical distributions do not match the actual distribution of giftedness in the population. Silverman suggested that with each new set of published norms, the problem intensifies as the scores of gifted children are compressed and lowered to fit the normal distribution, even taking into account the Flynn effect.

Although measurement of intelligence in gifted populations appears fraught with difficulty, the Stanford-Binet, Fifth Edition (SB5) was designed, in part, to specifically meet the needs of the gifted population. The authors (Roid, 2003) convened special meetings with experts in gifted education to discuss test design issues. In response to their feedback, the SB5 includes features intended to be useful for this population. There are items of high difficulty on the test so that extreme levels of giftedness can be documented. There is a decreased emphasis on timed tasks that can unfairly penalize gifted children who possess a reflective thinking style (Kaufman, 1994). The inclusion of a large age range makes it possible to test kindergarten-aged gifted children on the same test as older students and allows age-equivalent scores that extend well into adulthood.

Despite these attempts to develop a test useful for determining giftedness, there were several early indications that the test was not operating in the expected manner. The gifted sample included in the validity studies of the test (Roid, 2003) received a mean Full Scale score of 123.7--over 6 points below the general standard of 130 for gifted-program participants. In an additional study (Kearney, 2004) Riverside Publishing documented that students who were performing at a prodigious level received a mean score of 130 on the SB5, a score much lower than expected. In addition to the lower means, it seems odd that a sample made up primarily of moderately gifted students would only score 6 to 7 points below a sample in the profoundly gifted range.

In order to better understand the impact of the SB5, our local group of school administrators, parents, and school psychologists worked together to define several questions to be answered through this study. First, we wanted to know whether this test discriminated well between gifted and highly gifted populations using the current guidelines of 130 for gifted programs and 145 for highly gifted programs, assumed to be two standard deviations and three standard deviations from the mean and at the 98th and 99.9th percentile, respectively. We also wished to know whether the age equivalent scores and the experimental gifted composite score would be useful indicators of functioning at these two levels of giftedness.

Method

Study participants were elementary-school students in grades two through six from Boise, Idaho. Students met the requirements and had been admitted to one of three gifted program options within the Boise School District. These options were: a) a gifted pull-out program in which children attend enrichment classes one-half to one day per week, b) full-time classrooms for gifted children, or c) full-time classrooms for highly gifted children. Two levels of giftedness are currently used by the school district: gifted and highly gifted. To be designated as gifted, a student must meet three of five criteria. One of the five criteria must be number 1 or number 2 of the following:

1. Intelligence test score at or above the 98th percentile.

2. Achievement test scores at or above the 98th percentile in the majority of areas tested.

3. Two nominations indicating potential ability.

4. Results of creativity/critical thinking test at or above the 98th percentile.

5. Documentation of talent from a facilitator-recognized expert in the field.

In addition, there must be documentation that the student's needs cannot be provided for with regular classroom instruction on a full-time basis. Students with multiple exceptionalities are evaluated on an individual basis with consideration given as to how their exceptionality might impact test scores. To be designated as highly gifted, a child must meet similar criteria, but with an intelligence test score at or above 145. Students who did not receive a score of 145 on the WISC-III but who received several subscale scores that hit the "ceiling" on the WISC-III (defined as an 18 or 19 standard score on a subtest) were also given the option of testing on the Stanford-Binet Form L-M to meet the IQ requirement. It should be noted that virtually all students in the gifted programs met the IQ cut-off requirements specified by the district.

All highly gifted children were invited to be in the study. Out of 48 possible participants, 37 consented to be in the study. A total of 36 gifted children were recruited. These students were from seven schools to which participating school psychologists were assigned. The majority of the gifted children had admitting IQ scores between 130 and 134.

After receiving informed consent, each child was administered the SB5 by one of six experienced school psychologists between January and February, 2004. School psychologists had been trained in test administration through publisher workshop attendance, local in-service training, publisher consultation, and peer consultation. Hand-tabulated scores were verified using the SB5 Scoring Pro software. Analyses were completed using the Statistical Package for the Social Sciences (SPSS) and Microsoft Excel.

Results

Table 1 shows the means and standard deviations for the SB5 Full Scale IQ and the experimental Gifted Composite score for both gifted and highly gifted groups. The Full Scale SB5 scores did differ significantly between gifted and highly gifted children (Mann-Whitney U = 412.5,p < .01). Fifty-six of these participants had been previously tested on the WISC-III. The means and standard deviations of these scores are also presented in Table 1. The remainder of the participants had been tested on a variety of other instruments, primarily the Stanford-Binet-IV, The Stanford-Binet Form L-M, or the Wechsler Preschool and Primary Scale of Intelligence. Full Scale and Gifted Composite scores were significantly lower on the SB5 than on the Full Scale scores on the WISC-III (Wilcoxon Z = -6.513, p < .01 and Z = -6.485, p < .01, respectively).

Table 2 shows the means, standard deviations, minimum and maximum scores of gifted and highly gifted students on the Abbreviated IQ score, scores in the Verbal and Nonverbal Domains, and the five factors scores of Fluid Reasoning (FR), Knowledge (KN), Quantitative Reasoning (QR), Visual Spatial (VS), and Working Memory (WM).

The discrepancy between the SB5 Nonverbal and Verbal means in our sample is small--3 points for gifted and 3 points for highly gifted students. Ten students had a 12-point discrepancy or greater. This outcome is dissimilar to the discrepancy rates for WISC-III data, which show 23 students to have at least a 12-point discrepancy. Only 6 of those students also showed at least a 12-point discrepancy on the SB5--all in the same direction of cognitive preference with verbal scores being higher than nonverbal.

To address the question of whether the ranks are preserved between the two tests, measures of correlation were used. Since the assumption of normality can be questioned with scores this extreme, Kendall's tau b and Spearman's rho were calculated. The correlation between the SB5 and the WISC-III using Kendall's tan b was .341. Using Spearman's rho the correlation was .457. The correlation between the two tests for all IQ levels was .82 (Roid, 2003). To illustrate the type of variation found in the raw data, Table 3 shows the highest five ranks of the SB5 and the ranks of the scores each of these students achieved on the WISC-III. Table 4 shows the five lowest ranks on the SB5 and the ranks of the corresponding WISC-III scores. Table 5 shows the percentage of students who would be designated as non-gifted, gifted and highly gifted for two different levels of cut-off scores using the Full Scale IQ score and the experimental gifted composite score. Data relating to age-equivalent scores is shown in Table 6. These scores indicate a substantial portion of our students in both groups were operating at an age level much higher than their chronological age.

The SB5 Experimental Composite Score for Intellectual Giftedness correlated highly with the SB5 Full Scale IQ with a Pearson r of .95 (p < .01). The average gain of IQ in the gifted sample was .9 with scores ranging from a loss of 3 points from the Full Scale IQ to a gain of 5 points. The average gain in the highly gifted group was 1.6. Some highly gifted students lost as many as 6 points from their Full Scale IQs while others gained as many as 8 points.

Discussion

The means of our gifted and highly gifted groups on the SB5 are significantly lower than their scores on the WISC-III even after allowing a 5-point adjustment for the Flynn effect. The mean Full Scale score for our gifted group was 12 points lower than the corresponding WISC-III score, and the mean Full Scale score for our highly gifted group was 18 points lower than their previously earned WISC-III score. Using a strict cutoff of 145 for entrance to our highly gifted program means that not a single student would have qualified using the SB5 Full Scale IQ. Within our sample, 16 of 73 students would have qualified for gifted programs using a strict cutoff of 130 which is the criterion adopted by many districts as an indicator of giftedness. This is approximately 22% of our total sample, of which half had previously qualified as highly gifted. While one might expect some attrition in programs for the gifted when tests due to renorming and regression artifacts, the proportion of students affected using the SB5 seems excessive. The fact that children have also met other criteria for giftedness, including achievement standards, suggests that using the criterion of two or three standard deviations from the mean as an indicator of giftedness on this test is likely too high, even though this was the standard originally suggested by the publisher (Ruf, 2003). As this article was being written, Riverside Publishing was in the process of posting a suggested gifted cutoff that allows use of the Abbreviated Battery, either the Nonverbal or Verbal IQ alone, the Gifted Composite score, or a combination of ability domain scores to be valid in assessing giftedness. They suggest that even when a child's Full Scale score is around 120, the additional combination of other scores at the 95th percentile and above should be considered indicative of a need for services when viewed in combination with other indicators such as achievement, recommendations, and portfolios (D. L. Ruf, personal communication, April 5, 2005). This recommendation is consistent with our finding that using a guideline of 130 on this test to determine moderate giftedness is too high.

Although a complete discussion of regression artifacts is not practical here, it should be mentioned that whenever extreme scores are used on an instrument for classification purposes, there should be large amounts of regression toward the mean when a second test is given (Campbell & Kenny, 1999). Studies such as Milich, Roberts, Loney, and Caputo (1980) and Furlong and Feldman (1992) demonstrate how the regression artifact dramatically influences classification of hyperactivity and learning disabilities. Campbell and Kenny report that, given the assumption of normality and retest correlations of .80, the probability that a person who scores two standard deviations above the mean actually has a true score this high is .58. (This probability increases to .86 if the person has three test scores that are all two standard deviations above the mean.) Regression artifacts are undoubtedly influencing our data; however, estimating their effect is problematic given other confounding factors such as the use of two scales based on somewhat different theoretical models of intelligence.

The fact that the mean Full Scale IQ score was significantly different between our gifted and highly gifted students suggests that the SB5 does an adequate job of broadly distinguishing between these groups. Interestingly, the difference between the means of these groups on the SB5 is only 5 points; this is not even close to the 15 point standard deviation of the measure, and is less than half of the 11 point difference in groups found on the qualifying WISC-III. In addition, the lack of verbal-nonverbal discrepancy transfer from the WISC-III (41% of WISC-III takers) to the SB-V (8% of SB5 takers) suggests differences in the nature of capturing intellectual giftedness with the latter measure, although regression effects could also be operating here.

It is instructive to look at the range of the SB5 data for our gifted and highly gifted groups. The findings in Table 2 show that on every subscale score some students in both groups scored within the average range. Related to this finding is that the range of scores on each factor varied between 26 and 57, indicating large differences in how our students scored across various content areas. In our highly gifted group the range of scores on quantitative reasoning confirmed anecdotal information from teachers that this is the area in which their students show the most variation with some working at grade level, while others are capable of working well into high school and college material.

Another interesting finding is that some individual's scores varied dramatically between the two intelligence tests (as much as 35 points). Two students in the highly gifted program scored so low on the SB5 that they would not meet intelligence test criteria for giftedness even if the cut-off were lowered to 115. Alternatively, several students who scored at the low end of the IQ cut-off for our gifted program did very well on the SB5 and scored within the range of the study completed by Riverside on child prodigies. To check on whether these children were misclassified as gifted or highly gifted, anecdotal information was requested from their teachers. These data suggest that these children's original classifications were probably correct. For example, one of the highly gifted children had successfully completed a high-school level course as an elementary student; it is unlikely that a non-gifted child would perform in such a manner. It should be noted that in many cases, the SB5 did correctly distinguish between gifted and highly gifted children. However, the relatively low correlation between it and the WISC-III in our sample, suggests that these two measures may be capturing somewhat different abilities--at least in the gifted population.

In contrast to the means on the SBS, which seem to be lower than expected, many of the age-equivalent scores appear to be higher than one might expect. The proportion of children that received age-equivalent scores beyond the highest average score for mature adults (19% of gifted students and 40% of highly gifted students) seems implausible even after taking into account documented cautions in interpreting such scores (Sattler, 2001). If these age-equivalent scores are indicative of actual functioning, it seems curious that they are not accompanied by correspondingly high IQ scores. In sum, these scores appeared to be of minimal use in accurately describing levels of giftedness in our sample.

The high correlation between the experimental composite score for intellectual giftedness and the full scale IQ score indicates that the constructs measured are quite similar. With this in mind, administration of the gifted composite would represent a time savings for busy school psychologists, since they would only have to administer a portion of the test. It should be understood, however, that this practice would have implications for individual children. For example, those who perform well on Working Memory and Nonverbal Visual-Spatial Processing may achieve lower IQ scores than they would have if the Full Scale IQ was administered. Fifteen students in our gifted sample and 9 students in our highly gifted sample "lost" IQ points in this manner, though in many cases, the difference would not have influenced program eligibility. Whether this issue is considered a problem has much to do with conceptual definitions of giftedness and with decisions to emphasize certain intellectual aspects over others (e.g., reasoning vs. memory). When in doubt, the most prudent course may be to administer the entire test and examine the experimental composite within the interpretive process.

More data need to be collected with other groups of gifted and highly gifted children to see whether our findings can be generalized or whether they relate only to our sample of children. In a descriptive study such as this, it is impossible to determine how much of the mean score changes are accounted for by regression artifacts, the Flynn effect, actual conceptual differences in the definition of intelligence, and other factors. In general, we found that SB5 full scale and factor scores are significantly lower than would be expected based on previous IQ testing and even taking into account the Flynn effect. In addition, group mean discrepancies, as well as verbal-nonverbal discrepancy patterns, did not hold up well between the WISC-III and SB5. Whether these findings are due to statistical issues, variation in test content, or some oddity of our sample is beyond the scope of the current study. Using these data as guidelines and bearing in mind that the use of tests should show "the relative absence of adverse consequences" (Messick cited in Roid, 2003) for children that are truly gifted, we believe professionals should be wary of using rigid cutoff scores for admission to gifted programs with the SB5. Instead, they should use multiple criteria, with such scores being considered a single, but fallible, source of information. Because some gifted and highly gifted children are receiving relatively low scores on this test, students should not be excluded from gifted programs on the basis of such scores alone. Until we have more information about the use of this measure, we suggest that others carefully assess and adapt their gifted qualifying guidelines in conjunction with a comprehensive evaluation addressing such sources as academic scores, portfolios, parent reports, and additional talent measures. While we feel the SB5 may fit into such an overall assessment, its use, as with any IQ measure, should be tempered by appropriate queries about what is measured, how it is measured, and how its scores are distributed in the gifted population.

The authors would like to thank the gifted children who were willing to be retested for this study and their parents who allowed them to participate. We would also like to acknowledge the support of Jo Henderson, Supervisor of Gifted Services for the Boise School District and the six school psychologists who administered the tests.

Manuscript submitted December 8, 2004.

Revision accepted April 16, 2005.

REFERENCES

Campbell, D. T., & Kenny, D. A. (1999). A primer on regression artifacts. New York: The Guilford Press.

Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytical studies. New York: Cambridge University Press.

Flynn, J. R. (1984). The mean IQ of Americans: Massive gains 1932 to 1978. Psychological Bulletin, 101, 171-191.

Flynn, J. R. (2000). The hidden history of IQ and special education: Can the problems be solved? Psychology, Public Policy, & Law, 6, 191-198.

Furlong, M. J., & Feldman, M. G. (1992). Can ability-achievement regression toward the mean account for MDT discretionary decisions? Psychology in the Schools, 29, 205-212.

Kanaya, T., Scullin, M. H., & Ceci, S. J. (2003). The Flynn effect and U.S. policies: The impact of rising IQ scores on American society via mental retardation diagnoses. American Psychologist, 58, 778-790.

Kaufman, A. S. (1994). Intelligent testing with the WISC-III. New York: Wiley.

Kearney, K. (2004, November). Using the Stanford Binet Five with the gifted: Options for special schools and programs. Paper presented at the National Association for Gifted Children, Salt Lake City, UT.

Milich, R., Roberts, M. A., Loney, J., & Caputo, J. (1980). Differentiating practice effects and statistical regression on the Conners Hyperactivity Index. Journal of Abnormal Child Psychology, 8, 549-552.

Roid, G. H. (2003). Stanford-Binet intelligence scales, fifth edition, technical manual, Itasca, IL: Riverside Publishing.

Ruf, D. L. (2003). Use of the SB5 in the Assessment of High Abilities (Stanford-Binet Intelligence Scales, Fifth Edition Assessment Service Bulletin No. 3). Itasca, IL: Riverside Publishing.

Sattler, J. M. (2001). Assessment of children: Cognitive applications (4th ed.). San Diego, CA: Jerome M. Sattler.

Silverman, L. K. (1989, November). Lost: One IQ point per year for the gifted. Paper presented at the convention of the National Association for Gifted Children, Cincinnati, OH.

Terman, L. M. (1925). Genetic studies of genius: Vol. 1. Mental and physical traits of a thousand gifted children. Stanford, CA: Stanford University Press.

Wechsler, D. (1944). The measurement of adult intelligence. Baltimore, MD: Williams & Wilkins.

Wechsler, D. (1991). Wechsler intelligence scale for children--Third edition manual. San Antonio, TX: Psychological Corporation.

Wechsler, D. (2003). Wechsler intelligence scale for children Fourth edition: Administration and scoring manual. San Antonio, TX: The Psychological Corporation.

Barbara Minton has a joint Ph.D. in developmental and child clinical psychology. In addition to homeschooling a profoundly gifted child, she assesses gifted children and advocates for the development of services to meet their needs. E-mail: [e-mail]

Shannon Pratt, Ph.D., NSCP, is a school psychologist. He has three young children (one of whom is a gifted reader). His interests are in gifted assessment, behavioral interventions, and the evaluation of behavior disorders. E-mail: [e-mail]

Table 1
Means and Standard Deviations of Gifted and
Highly Gifted Students on the SB5,
the SB5 Experimental Composite Score for
Intellectual Giftedness, and the WISC-III

                                               Highly
                              Gifted           gifted
                             (n = 36)         (n = 37)

SB5 full scale IQ          Mean     121     Mean     126
                           SD       6.4     SD       7.8

SB5 gifted composite       Mean     122     Mean     128
                           SD       6.8     SD       9.1

                                               Highly
                              Gifted           gifted
                             (n = 25)         (n = 31)

WISC-III full scale IQ     Mean     133     Mean     144
                           SD       3.6     SD       6.3

Table 2
Means, Standard Deviations,
Minimum and Maximum Scores of
Gifted and Highly Gifted Students on
Factor Scores, Verbal and Nonverbal
Domain Scores, and Abbreviated IQ

                           Mean (SD)      Minimum     Maximum

Abbreviated IQ
  Gifted                   116 (8.8)         94         133
  Highly gifted            121 (7.5)         97         150

Nonverbal IQ
  Gifted                   119 (7.0)        104         136
  Highly gifted            124 (9.4)        100         141

Verbal IQ
  Gifted                   122 (6.9)         98         134
  Highly gifted            127 (7.5)        110         138

Fluid reasoning
  Gifted                   118 (9.3)         88         132
  Highly gifted            125 (9.8)        109         141

Knowledge
  Gifted                   118 (7.6)        100         131
  Highly gifted            123 (8.7)        100         140

Quantitative reasoning
  Gifted                   121 (9.8)        103         136
  Highly gifted            124 (13.3)        92         149

Visual spatial
  Gifted                   121 (9.8)        103         146
  Highly gifted            123 (10.2)       103         149

Working memory
  Gifted                   115 (7.5)        103         129
  Highly gifted            119 (7.5)        106         138

Table 3
Stanford-Binet, Fifth-Edition Scores of Top
Ranking Gifted Students on the WISC-III

Rank on WISC-III     WISC-III IQ     Rank on SB5     SB5 IQ

1                        154             10           128
1                        154             18           120
2                        153              2           138
3                        152             10           128
4                        150              3           136
5                        149             24           114

Table 4
Stanford-Binet, Fifth-Edition Scores of
Lowest Ranking Gifted Students
on the WISC-III

Rank on WISC-III     WISC-III IQ     Rank on SB5     SB5 IQ

23                       127             26           110
22                       130             23           115
                         130             21           117
                         130             20           118
                         130             18           120
                         130             17           121
                         130             13           125
                         130             12           126
21                       131             27           102
                         131             25           112
                         131             11           127
                         131              9           129
20                       132             20           118
19                       133             24           114
                         133             23           115
                         133             21           117
                         133             19           119
                         133             17           121
                         133             15           123
                         133             15           123
                         133             11           127

Table 5
Classification of Students As Gifted
or Highly Gifted Using Two Cut-off
Scenarios On the SB5

Original                             Gifted         Full
designation                         composite     scale IQ

Using cut-offs of 115 (gifted) and 125 (highly gifted)
  Gifted students
    Identified as non-gifted         4 (11%)       5 (14%)
    Identified as gifted            20 (56%)      20 (56%)
    Identified as highly gifted     12 (33%)      11 (31%)

  Highly gifted students
    Identified as non-gifted         4 (11%)       3 (8%)
    Identified as gifted             9 (24%)      12 (32%)
    Identified as highly gifted     24 (65%)      22 (59%)

Using cut-offs of 120 (gifted) and 130 (highly gifted)
  Gifted students
    Identified as non-gifted        12 (33%)      13 (36%)
    Identified as gifted            12 (33%)      20 (56%)
    Identified as highly gifted     12 (33%)       3 (10%)

  Highly gifted students
    Identified as non-gifted         7 (19%)       7 (19%)
    Identified as gifted            12 (32%)      16 (43%)
    Identified as highly gifted     18 (49%)      14 (38%)

Table 6
Number of Years Mental Age is
Ahead of Chronological Age Using
Age Equivalent Scores for Gifted and
Highly Gifted Students on the SB5

                                  Number of students

Years ahead                    Gifted     Highly gifted

1                                1              0
2                                2              2
3                                5              3
4                                5              4
5                                5              1
6                                3              1
7                                2              3
8                                0              2
9                                2              1
10                               2              2
11                               2              3
Beyond the highest average       7             15
  score for mature adults