A Review of the Reynolds Intellectual Assessment Scales,Second Edition,and Reynolds Intellectual Screening Test,Second Edition

Administration, Scoring, and Interpretation

The RIAS-2’s four intelligence subtests (Guess What, Odd-Item Out, Verbal Reasoning, and What’s Missing) can be administered to most examinees in about 25 min. The two memory subtests (Verbal Memory and Nonverbal Memory) can be administered in about 10 min, and the two processing speed subtests (Speeded Naming Task and Speeded Picture Search) can also be administered in about 10 min. The RIST-2 can typically be administered in about 12 min. Individuals with adequate training can easily administer and score the RIAS-2 subtests. Users of the RIAS (Reynolds & Kamphaus, 2003) will find that the discontinue rules on verbal and nonverbal subtests now include one and two additional items, respectively, so these subtests will take more time to administer than the previous edition.

The RIAS-2 kit contains a Fast Guide that appears to have enough information to administer the RIAS-2 in lieu of the Professional Manual. The RIAS-2 Record Form booklet is 28 pages in length and includes instructions, prompts, examples, and test items in an easily read layout for all eight subtests. Stimulus books are sturdy yet easy to manipulate, and their pages contain multi-colored complex images and two-toned shapes. No manipulatives are required to administer the RIAS-2, but unlike the RIAS, one subtest requires use of a pencil or pen.

The eight RIAS-2 subtests yield raw scores and T-scores (M = 50, SD = 10). Subtests yield five composites: the Verbal Intelligence Index (VIX), the Nonverbal Intelligence Index (NIX), the Composite Memory Index (CMX), the Speeded Processing Index (SPI), and the Composite Intelligence Index (CIX). The SPI is new to the RIAS-2. The VIX, NIX, CMX, and SPI stem from two subtests, and the CIX stems from the four VIX and NIX subtests. The RIAS-2 authors also offered information to support the use of three additional composites scores, Total Verbal Battery (stemming from four subtests), Total Nonverbal Battery (stemming from four subtests), and Total Test Battery (stemming from all eight subtests), in an appendix. The RIST-2 produces only a single composite called the RIST-2 Index that stems from two subtests. All composite scores are reported as deviation IQ scores (M = 100, SD = 15), and additional scores can also be obtained. For composite scores, confidence intervals (based on the standard error of estimate) surround estimated true scores.

Development

Development of the RIAS-2 began with a market research survey to determine what revisions were desired by users and was guided by nine goals cited in the its Professional Manual. Even further, the RIAS-2 authors provide a practical and empirical rational for the development of this test based on previous research and review of the literature. Similar to the RIAS, the RIAS-2 was developed with the goals of providing a sound measure of general intelligence that accurately predicts academic achievement across a broad age range, while eliminating or reducing examinee reading and motor coordination requirements, as well as reducing bias in testing by using a simple, brief, and easy administration format. A pilot test was conducted to evaluate RIAS-2 item characteristics, and a variety of item-level analyses (based on both classic test theory and item response theory) were completed. An item bias evaluation was conducted by five individuals from diverse backgrounds, and differential item functioning associated with gender and ethnicity was evaluated.

Standardization and Norming

Norms for the RIAS-2 included 2,154 individuals based on a population-proportionate, stratified random sampling plan using the 2012 United States Census population statistics. Normative data were collected from February 2013 to August 2014 across 32 states. No Canadian norms are available. Norms were organized into 52 norm blocks ranging from 4-month intervals to 10-year intervals; there were approximately 30 to 96 individuals associated with each norm block from the ages of 3 to 7 years and 13 to 94 years, which meets the minimum standard for size (Norfolk et al., 2015). However, the norm blocks within the age ranges of 8 to 12 years were estimated to contain 29 to 29.6 norm sample participants per block, which falls just below the minimum standard. Despite these limitations, continuous norming and sample weighting procedures were used.

Scale and Item Characteristics

There is no evidence of ceiling violations for subtests. Thus, the maximum raw score was always associated with a norm-referenced score that is at least two standard deviations above the mean. Two subtests demonstrated floor violations within the range of ages 3:0 to 3:11. Thus, there are no serious concerns about the RIAS-2’s ability to measure high and low levels of abilities at age 4:0 and older.

Reliability

For every age group analyzed, coefficient alpha values for subtests were .80 or higher (Mdn across age groups = .86). These findings were generally consistent across gender and race and ethnicity (White, African American, and Hispanic). RIAS-2 composites demonstrated higher reliability values than subtests (Mdn across age groups = .92). They were typically .90 or higher, with a few exceptions (primarily, the NIX and CMX composites). Notably, coefficients for CIX ranged from .91 to .97 across age groups (Mdn = .93), and the same pattern of high reliability was evident across the Total Battery composites. Coefficients for the RIST-2 Index ranged from .88 to .95 across age groups (Mdn = .92). Reliability values for subtests and composites were generally consistent across genders and racial and ethnic groups.

Median test–retest reliability coefficients for subtests (across 7 to 43 days, with a median interval of 18 days and corrected for attenuation) exceeded .80 across four age groups in all but two cases (Nonverbal Memory and Speeded Picture Search). Furthermore, corrected test–retest reliability coefficients well exceeded .90 for the CIX, VIX, Total Test Battery, Total Verbal Battery, and RIST-2 Index composites for every age group (with a single exception); median values for NIX, CMX, SPI, and Total Nonverbal Battery exceeded .80 across age groups (with a single exception). There is no evidence of the stability of RIAS-2 scores beyond approximately 1 month. Interscorer reliability was assessed by two members of the publisher’s staff independently scoring 35 completed protocols; resulting correlations were perfect or near perfect. The magnitude of these reliability coefficients as a whole provides evidence of the consistency with which the RIAS-2 items, subtests, and composites measure targeted abilities and generally supports the assertion that composites rather than subtests should be interpreted.

Validity

The RIAS-2 authors addressed validity evidence organized according to test content, response processes, internal structure, relationship with other variables, and consequences of testing. In particular, they reported the results of exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) using the norming sample data from all eight subtests. One-factor solutions, representing general intelligence, indicate that the two VIX subtests, Guess What and Verbal Reasoning, are strongly correlated with the general factor (with g-loading values at or exceeding .70), whereas the two NIX subtests, Odd-Item Out and What’s Missing, have moderate g loadings (within the range of .50 to .69). By design, the CMX and SPI subtests all demonstrated weak g loadings (as low as .04 for Speeded Picture Search). Two- and three-factor solutions (using only an orthogonal rotation) from EFA did not indicate the viability of the four-factor structure of the RIAS-2. Instead, they indicate that one memory subtest (forming the CMX) measures the same factor as the VIX subtests and the other memory subtest measures the same factor as the NIX subtests. These same patterns appeared evidence across analysis by gender and race/ethnicity. No four-factor solution from an EFA was offered. In general, EFA results supported findings from research using the RIAS with referred and clinical samples (e.g., Dombrowski, Watkins, & Brogan, 2009; Nelson & Canivez, 2012) that indicated concerns about the validity of the RIAS CMX in measuring the memory construct.

A total of 14 models (representing one-factor and correlated two-, three-, and four-factor models using four to eight RIAS-2 subtests) were specified and evaluated using CFA. When considering fit statistics and parameter estimates for models using all eight subtests, the three-factor model specifying Verbal, Nonverbal, and Processing Speed factors (with the two CMX subtests loading on their respective Verbal and Nonverbal factors) was best fitting. The four-factor model consistent with the structure of the four RIAS-2 composites included out-of-bound values and demonstrated worse fit. Although no hierarchical EFA models or higher-order CFA models were evaluated, results from factor analysis support the interpretation of the CIX, VIX, NIX, and SPI composites but undermine the confidence in interpreting the CMX (as its subtests seem to measure different latent variables) and the three Total Battery composites (as the SPI subtests do not seem to measure the same latent variables as do the CIX, VIX, and NIX).

Copious validity evidence based on external relations was presented. For example, RIAS-2 composite scores were correlated with subtest scores and composite scores from four other measures of intelligence. Results were generally as expected. One exceptional finding, however, was that the SPI correlated with other composites measuring Processing Speed no more strongly (and sometimes more weakly) than several other composites. Two measures of achievement and one measure of memory were correlated with RIAS-2 scores, and correlations between the RIAS-2 CIX and achievement composite scores were moderate in magnitude. Additional evidence of external relations stems from 12 specific clinical groups, including intellectual disability, traumatic brain injury, learning disability, and attention-deficit/hyperactivity disorder. Score differences across clinical and matched control groups were uniform as expected.