The Cognitive Abilities Scale—Second Edition (CAS-2; Bradley-Johnson & Johnson, 2001) is the second edition of a test that was first published in 1987. It is an individually administered, norm-referenced test designed to assess cognitive abilities in children ages 3 through 47 months. The authors' goal in developing the scale was to "assess current level of functioning and to identify children who would benefit from special instruction in order to improve their abilities" (p. 2). In order to enhance the scale's use in educational decision-making, the authors sought to 1) use items that measure early childhood abilities that tend to predict future performance, and 2) create tasks that are similar to those used in a classroom.

The CAS-2 is not based on a single theoretical approach. Rather, the authors based the scale on current theories (information processing, learning, and visual perception), Piaget's work, and relevant research on early cognitive development.

The CAS-2 is composed of two forms. The Infant Form (3 months through 23 months) samples abilities in three areas: exploration, communication, and initiation and imitation. The Preschool Form (24 months through 47 months) samples abilities in five areas: oral language, reading, mathematics, handwriting, and enabling behaviors. Both forms yield two quotients: the General Cognitive Quotient (GCQ) which measures cognitive ability on all items, and the Nonvocal Cognitive Quotient (NCQ) which measures cognitive ability on the non-vocal items alone. The NCQ is a feature that allows for a reliable and valid measure of children who are unwilling or unable to speak during testing or whose speech cannot be understood.

GCQ's and NCQ's are obtained by converting raw scores to a normalized standard score with a mean of 100 and a standard deviation of 15. Tables for converting quotients into percentiles and age equivalents are also provided. The authors provide explicit instructions for calculating a 68% confidence interval using the standard errors of measurement (SEM) associated with age and quotient of interest. Explicit instructions for calculating alternative confidence intervals (e.g. 90% and 95%) are not provided, with the authors assuming those with adequate background in statistics can calculate these given the SEM of interest.

Infant Form (3-23 months)

Exploration of Objects (25 items)—These items assess an infant's ability to explore their surroundings using vision, hearing, and touch. Items assess skills such as reaching, and grasping and abilities such as distinguishing cause-effect relationships and problem-solving.

Communication with Others (25 items)—These items assess an infant's early receptive and expressive communication abilities. Items measure behaviors such as independent production of vocal sounds, response to vocal sounds, imitation of vocal sounds, use of gestures, and ability to follow directions.

Initiation and Imitation (29 items)—These items assess "abilities that enable children to learn" (p. 8). They measure an infant's ability to initiate or model skills such as book handling, using a paper and pencil, and sequencing objects based on size. Items also measure an infant's ability to imitate specific behaviors when requested to do so by the examiner.

Preschool Form (24-47 months)

Oral Language (30 items)—These items assess a child's ability to both comprehend and use oral language. Language skills measured include understanding of position words, expressive and receptive knowledge of nouns and pronouns, and ability to use plurals, word endings, articles, and correct word order when speaking.

Reading (16 items)—These items measure early literacy skills such as book handling, comprehension, letter naming, and providing letter sounds.

Mathematics (22 items)—These items measure early mathematic skills. Items assess a child's ability to arrange objects by size, match numbers with quantities, find matching sets, understand general mathematical concepts (e.g. little or more), recognize numbers, and count.

Handwriting (6 items)—These items assess abilities used in manuscript writing such as pencil grip, posture, and the ability to copy figures involved in letter formation.

Enabling Behaviors (14 items)—These items assess abilities that aid effective learning. Items measure a child's ability to imitate behaviors, remember sentences, and remember lists of unrelated words.

Norms

1,106 children from 27 states were tested for the CAS-2 normative sample. These children were tested from October 1997 to August 1999. Primary sites from all four major geographical regions in the U.S. were chosen in order to standardize the CAS-2. The primary sites consisted of daycare or preschool programs, school districts, and medical centers. No information regarding how many children were selected from each site is provided. The authors compare the normative sample to the U.S. population under the age of 5 (according to the 1997 census) according to the following demographic characteristics: geographic area, gender, race, residence, ethnicity, educational attainment of parents, and disability status. The norms appear representative of the 1997 U.S. population across these characteristics, and the authors make special mention of the fact that the percentage of children with disabilities in their sample is also representative of what would be expected in the U.S. population. However, crosstabulations are shown only for age by geographic region, gender, race, residence, and ethnicity. Thus, we do not know whether, for example, whether all African-American children were male, from the Northeast, and so on. The norm sample appears representative of the U.S. population when crosstabs based on age are provided.

Reliability

The authors report information on internal consistency, test-retest, and interrater reliability.

Internal consistency was measured using Cronbach's (1951) coefficient alpha. Coefficient alphas are reported for both Preschool and Infant GCQ's and NCQ's across age intervals and as an average across all ages. Coefficient alphas for age intervals range from .84 to .96. Average coefficient alphas range from .89 to .94. The majority (68%) attain the .90 criterion for making decisions about individual students. In order to demonstrate reliability across subgroups, coefficient alphas are also provided for five subgroups (male, female, European American, African American, and Hispanic American) within the norm sample. These alphas are all equal to or greater than .93.

Test-retest reliability was calculated for the Infant Form. During data collection for the Infant Form of CAS-2, 24 infants (ages 3-11 months) were re-tested a week after initial testing, and 21 infants (ages 13-22 months) were re-tested two weeks after initial testing. Test-retest correlations for the Infant Form GCQ were .81 and .90, and for the Infant Form NCQ were .79 and .93. No test-retest reliability was provided for the Preschool Form on the new standardization sample. The authors claim that the evidence for the CAS test-retest reliability is sufficient. Since the items and scoring criteria for the Preschool Form of CAS-2 and the original CAS are identical, data samples gathered for CAS were used to demonstrate test-retest reliability for the Preschool Form in CAS-2. Using two separate samples, in total 70 children (ages 2-3) were re-tested two weeks after initial testing. Test-retest correlations for the Preschool Form GCQ were .96 and .99, and for the NCQ were .98 and .92.

Interscorer agreement was also determined. Seventy-nine children were administered the CAS-2 while the examiner and an observer independently scored their performance. Coefficient alphas were found for the GCQ across four different age intervals. All alphas exceeded .95.

Validity

Content Validity The authors argue that content validity is demonstrated in a number of ways. First, the authors describe each area selected, and provide the research and rationale for including each item in the test. While the authors believe the test "provides valuable insights and useful information for planning instruction for children" (p. 32), they also note that ongoing research in this area is needed to verify the utility of selected areas and items.

Second, a differential item functioning analysis was also conducted in order to detect any bias in test items. Several items were found to be statistically biased (p< .001) for some groups of students. The authors reviewed these items and claim no bias is present.

Criterion-Related Validity

Evidence for criterion-related validity is provided based on three studies. Scores on both forms of the CAS-2 were compared to scores on the Bayley Scales of Infant Development-Second Edition (BSID-2; Bayley, 1993). Correlations ranged from .62 to .86, with lower correlations being found for the CAS-2 Infant Form. Scores on the CAS-2 Preschool Form were also compared to the Pictorial Test of Intelligence-Second Edition (PTI-2, French 2001) and the performance subtests of the Weschler Preschool and Primary Scale of Intelligence-Revised (WPPSI-R; Weschler, 1989). Correlations ranged from .67-.87.

No evidence is provided that the Infant Form predicts later cognitive abilities. Evidence for the predictive validity of the Preschool Form of CAS-2 is inferred from two studies comparing CAS (identical to the Preschool Form of CAS-2) data to scores on intelligence and achievement tests administered to the same children approximately 5 years later.

Construct Validity

The authors provide evidence for construct validity by establishing how student performance on CAS-2 follows patterns that would be expected if the test measured what it is intended to measure. 1) Since cognitive abilities are based on development, it would be expected that performance on CAS-2 would increase with age. CAS-2 mean scores (i.e. performance) increase with age (coefficients range from .70 to .91). 2) CAS-2 differentiates between groups with known discrepancies in cognitive abilities. Children diagnosed with mental disabilities had lower mean scores than those diagnosed with physical impairments. 3) The authors also highlight a study in which high correlations (r = .63-.84) were found between the CAS-2 Preschool Form and other achievement tests measuring academic abilities (i.e. TERA-2 and TEMA-2). 4) Using data provided earlier to demonstrate content validity as evidence for construct validity, the authors highlight the strong item discrimination correlations (.27-.42). Using evidence for content and criterion validity as evidence for construct validity is not appropriate. Yet, the first two sets of studies done do support the authors' argument for construct validity

Summary

The CAS-2 is a norm-referenced, individually administered test designed to assess cognitive abilities in children ages 3 to 47 months. The CAS-2 has two forms. The Infant Form (3-23 months) consists of three subscales: exploration of objects, communication with others, and imitation and initiation. The Preschool Form (24-47 months) consists of five subscales: oral language, reading, mathematics, handwriting, and enabling behaviors. The scales' norms appear representative in terms of geographic area, gender, race, residence, ethnicity, educational attainment of parents, and disability status. Crosstabs are provided only for age. Raw scores are converted to a normalized standard score with a mean of 100 and a standard deviation of 15. Both forms yield two standard scores: General Cognitive Quotient (GCQ) and Nonvocal Cognitive Quotient (NCQ). Scores can also be converted into percentiles and age equivalents.

There is good evidence for the technical adequacy of the CAS. Internal consistency for the GCQ for the Infant Form is adequate for making individual decisions across all but one age interval, and the GCQ reliability for the Preschool Form is adequate for making important decisions across all ages. The NCQ is less reliable than the GCQ (5 out of 11 correlations reach criterion). Scores were moderately stable for the Infant Form. No stability studies were reported for the Preschool Form as the authors provided evidence from stability research on the original CAS. Interscorer agreement is adequate for both forms. Evidence for the validity of the scale is adequate, though more studies should be conducted on the concurrent and predictive validity of the preschool form. The scale seems most useful in the terms identifying cognitive strengths and weaknesses and targeting areas for intervention.

(CAS predictive criterion related validity) In the first study, 20 children (ages 2-0 through 3-10) given the CAS were later tested on the Standford-Binet Intelligence Scale-Fourth Edition (SB-IV; Thorndike, Hagen, Sattler, 1986), Test of Early Reading Ability-Second Edition (TERA-2; Reed, Hresko, Hammill, 1989) and the Test of Early Mathematics Ability (TEMA; Ginsburg & Baroody, 1983). In the second study, 30 children (ages 2-0 to 3-11) assessed on the CAS were later assessed using the Weschler Intelligence Scale for Children-Revised (WISC-R, Weschler, 1974) and the reading and math sections of the Kaufman Test of Educational Achievement (KTEA; Kaufman and Kaufman, 1985).

CAS correlation to achievement and intelligence tests

SB-IV	.66
WISC-R	.63
TERA-2	.52
TEMA	.59
KTEA Reading Composite	.52
KTEA Math Composite	.50