Validity and Reliability of the CCAT

The Cerebrals Cognitive Abilities Test (CCAT) is the original three-subtest crystallized-intelligence battery developed in the Cogn-IQ research program. It contains Verbal Analogies (VA), Mathematical Problems (MP), and General Knowledge (GK), and was designed to measure general crystallized intelligence and scholastic ability in adolescents and adults. The CCAT was subsequently revised — most notably with the addition of a stopping rule for administration efficiency — and renamed the Jouve Cerebrals Crystallized Educational Scale (JCCES). The validation findings reported here apply to the original CCAT and establish the empirical foundation that the JCCES inherits and extends.

The validation framework

A new psychometric instrument needs three things to be considered trustworthy: high reliability (consistency of measurement), strong criterion-related validity (correlation with established measures of the same construct), and well-calibrated norms (a reference distribution against which scores can be interpreted). This study evaluated all three for the CCAT on a sample of 584 respondents aged 12 to 75, with norm establishment based on a 160-participant subsample compared against established cognitive batteries.

The validation comparison instruments were chosen to span the relevant construct space. The Reynolds Intellectual Assessment Scale (RIAS; Reynolds & Kamphaus, 2003) is a contemporary individually-administered intelligence test with separable verbal and nonverbal indices; the Scholastic Assessment Test, recentered version (SAT I; College Board) is the dominant US college-admissions test with broad coverage of verbal and quantitative academic skills; the Wechsler Adult Intelligence Scale, Third Edition (WAIS-III; Wechsler, 1997) is the most widely-used adult IQ battery, providing a Full-Scale IQ and separate Verbal/Performance composites. Strong correlations across all three would establish that the CCAT measures something close to general crystallized intelligence and broad academic readiness.

Reliability findings

Internal-consistency reliability was assessed using the Spearman-Brown corrected split-half coefficient — splitting the test items in half, computing the correlation between the two halves, and adjusting upward for the fact that the full test has twice as many items as either half (Cronbach, 1951, summarizes the methodological foundation). The Spearman-Brown correction is a standard approach for split-half reliability that produces results comparable to Cronbach’s alpha for tests of moderate length.

The full CCAT demonstrated excellent reliability with a Spearman-Brown corrected split-half coefficient of .97. The Verbal Ability scale (combining VA and GK subtests, omitting MP) achieved .96. Both values fall comfortably above the .90 threshold conventionally used for tests intended to support individual interpretive decisions. The standard error of measurement (SEM) for the full CCAT was 2.77, indicating that a respondent’s reported score is precise to within approximately ±2.77 points (or ±5.4 points at the 95% confidence interval).

The reliability values place the CCAT at the upper end of what crystallized-IQ batteries of comparable length typically achieve. The JCTI nonverbal-reasoning test, also from Cogn-IQ, produces alpha values in the .92-.96 range across age groups; the WAIS-III FSIQ has split-half reliability of approximately .98 across its 14-subtest battery. The CCAT achieves comparable reliability with a much shorter three-subtest format, partly because the items are well-targeted at the construct and partly because the test length is sufficient to support stable internal-consistency estimates.

Criterion-related validity findings

Validity was assessed by correlating CCAT indexes with the established comparison batteries. Across all comparisons, the CCAT showed strong positive correlations consistent with measuring general crystallized intelligence and academic readiness:

• RIAS Verbal Index: r = .89. The CCAT and the RIAS verbal-side composite share approximately 79% of their variance, which is the typical level of overlap between two well-validated crystallized-IQ measures.
• SAT I Composite: r = .87. Similar to the JCCES-SAT correlations reported in the broader JCCES validation study, which found r = .814–.858 across SAT versions.
• WAIS-III Full-Scale IQ: r = .92. Very strong correlation, indicating that the CCAT measures essentially the same construct that the most-validated adult IQ battery measures.
• WAIS-III Verbal IQ: r = .89. Comparable to the WAIS-III FSIQ correlation, reflecting that the verbal subtests of the WAIS-III dominate the CCAT–WAIS overlap.

The pattern of correlations supports several interpretations. First, the CCAT measures general crystallized intelligence as captured by the RIAS Verbal Index and WAIS-III Verbal IQ — the strongest correlations are with verbal-side measures, consistent with the CCAT’s verbal/knowledge construct definition. Second, the CCAT correlates with overall cognitive ability as captured by WAIS-III FSIQ at r = .92, suggesting that crystallized intelligence shares substantial variance with general cognitive ability — consistent with the broader cognitive-psychology literature documenting strong intercorrelation among cognitive batteries (Carroll, 1993). Third, the CCAT correlates with broad academic-achievement measures (SAT I) at levels typical of crystallized-IQ batteries.

Norm establishment

Norms for the CCAT were developed on a 160-participant subsample of the validation study. The procedure: CCAT scores were anchored against RIAS Verbal Intelligence Index (VIX) and WAIS-III FSIQ/VIQ values for the same respondents, producing equated CCAT scales that can be interpreted on the standard IQ metric (mean = 100, SD = 15). RIAS VIX adjustments over time were applied to ensure the resulting norms reflected the most current population reference distribution available.

The age-referenced structure permits CCAT scores to be interpreted within age bands across the 12-to-75 range covered by the validation sample. For applied use, this means a CCAT score of, say, 115 indicates the respondent’s performance is approximately one standard deviation above the mean for their age band, with the interpretation calibrated to the equivalent score on the well-established RIAS and WAIS-III batteries.

From CCAT to JCCES

The CCAT was subsequently revised and renamed the JCCES. The most substantial revision was the addition of a stopping-rule administration: testing terminates after a respondent makes a specified number of consecutive errors, improving efficiency without sacrificing measurement precision. The JCCES General Knowledge subtest item analysis documents how the stopping rule (five consecutive errors) was calibrated to the empirical difficulty distribution of the items. Other JCCES revisions involve item-pool adjustments and updated norms.

The JCCES inherits the CCAT’s reliability and validity foundation while adding administrative efficiency. Subsequent JCCES validation studies — JCCES vs GAMA factor analysis, JCCES vs ACT factor analysis, JCCES vs RIAS Verbal Scale PCA, JCCES CEI vs multiple cognitive and academic measures — extend the validation evidence to the revised instrument and confirm that the original CCAT findings reproduce on the JCCES.

This historical lineage matters for current users. The CCAT name still appears in older research literature; the JCCES is the current designation. Both refer to the same three-subtest crystallized-intelligence battery, with the JCCES representing the post-revision version with stopping-rule administration. The validation findings reported here apply to the original CCAT and form the empirical foundation that subsequent JCCES studies build on.

Practical implications

For researchers using the JCCES (or the older CCAT designation in legacy data): the test functions as a strong proxy for general cognitive ability, particularly the crystallized side, with reliability and validity at levels comparable to commercial cognitive batteries. The split-half reliability of .97 supports individual-level interpretation; the validity correlations (.87–.92 with established benchmarks) support the construct-validity case.

For test users in clinical or applied contexts: the CCAT/JCCES is not a replacement for full-battery cognitive assessment when a comprehensive multi-domain profile is needed. The three-subtest design covers crystallized intelligence well but does not directly assess fluid intelligence, working memory, processing speed, or other domains that full batteries like the WAIS or WISC cover. For research applications where general cognitive ability is a covariate or moderator, the CCAT/JCCES is a defensible choice; for clinical evaluations requiring a complete cognitive profile, it is one of several instruments to use rather than a stand-alone solution.

For test development methodology: the CCAT validation illustrates the empirical structure of a defensible new psychometric instrument — strong internal consistency, strong criterion-related validity across multiple comparison batteries, age-referenced norms anchored against established standards. The pattern is what construct-validity work looks like when done well, and the CCAT/JCCES program is a documented example of executing it across multiple sequential studies.

Methodological caveats

The 584-participant sample, while substantially larger than typical Cogn-IQ research samples, was a convenience sample drawn from voluntary online completers. Self-selection bias and demographic non-representativeness are recurring concerns; replication in more population-representative samples would strengthen the generalization. The norm subsample of 160 is small for establishing population reference distributions, and the norms are likely best understood as preliminary for the original CCAT until larger normative studies are conducted on the JCCES.

The split-half reliability is the dominant reliability statistic in this study. Test-retest reliability — whether scores are stable across repeated administrations — was not assessed and is an important complement to internal-consistency estimates. Modern reliability methodology also points to distributional bounds on Cronbach’s alpha and related indices that should be reported alongside the empirical estimates; this analysis predates that methodological convention.

The criterion-related validity correlations are based on the same sample for both the CCAT and the comparison batteries, which is the standard validation design but introduces some risk of common-method variance inflating the correlations. Independent-sample replication of the validity correlations would address this concern.

Frequently Asked Questions

What is the relationship between the CCAT and the JCCES?

The CCAT is the original three-subtest battery (Verbal Analogies, Mathematical Problems, General Knowledge); the JCCES is the revised version with stopping-rule administration and updated items. They measure the same construct — general crystallized intelligence — and the CCAT validation findings reported here form the empirical foundation that the JCCES inherits.

What does a split-half reliability of .97 mean?

It means that splitting the test items into two halves and correlating the scores produces an extremely high correlation, with the Spearman-Brown correction adjusting for the fact that each half is shorter than the full test. A value of .97 indicates that the two halves measure essentially the same thing, and that random measurement error contributes only about 3% to the variance of total scores.

How do CCAT correlations compare to other crystallized-IQ batteries?

The CCAT’s correlations with WAIS-III FSIQ (r = .92), WAIS-III VIQ (r = .89), RIAS Verbal Index (r = .89), and SAT I (r = .87) are at the upper end of what crystallized-IQ batteries typically achieve. Two well-validated IQ instruments measuring the same construct correlate at r = .80–.95; the CCAT is in the .85–.92 range, consistent with measuring the same general construct.

Why use a Spearman-Brown corrected split-half coefficient?

It is one of several methods for assessing internal-consistency reliability. The Spearman-Brown correction adjusts the half-test correlation for the fact that the full test is twice as long as either half; this produces a value comparable to (but not identical to) Cronbach’s alpha. For tests of moderate length with reasonably uniform item difficulty, the two methods give similar results.

Are the original CCAT norms still applicable to the JCCES?

The CCAT norms apply to the original CCAT. The JCCES revisions — particularly the stopping rule and any item-pool changes — warrant updated norms specific to the revised instrument. Subsequent JCCES validation work has established equivalence with established benchmarks (RIAS, WAIS, SAT), but for current applications the JCCES-specific normative work is the appropriate reference.