The construct validity of a cognitive test rests on the pattern of its correlations with other measures of the same and related constructs. A test that claims to measure crystallized intelligence should correlate strongly with established crystallized-IQ measures, more weakly with measures of distinct constructs (e.g., processing speed, working memory), and predictably with academic-achievement measures that draw on similar cognitive substrates. The Jouve Cerebrals Crystallized Educational Scale (JCCES) Crystallized Educational Index (CEI) was tested against this validation pattern using a multi-instrument convergent study design: respondents who had completed the JCCES were also tested on, or self-reported scores from, a wide range of established cognitive and academic measures. The resulting Pearson correlations are substantial and consistent with the JCCES CEI being a valid index of general cognitive ability and academic achievement across age groups.
The validation framework
Crystallized intelligence — the accumulated knowledge and skills shaped by education and experience (Horn & Cattell, 1966) — is one of the most replicated constructs in cognitive psychology. It correlates strongly with verbal-loaded subtests of major IQ batteries, with academic-achievement measures, and with vocational-success indicators (Deary, Strand, Smith, & Fernandes, 2007; Neisser et al., 1996). A new test of crystallized intelligence should reproduce this correlation pattern; if it does, the convergent validity argument supports the test as a member of the crystallized-IQ family.
The JCCES is a three-subtest battery (Verbal Analogies, Mathematical Problems, General Knowledge) administered online. The Crystallized Educational Index (CEI) is the composite score derived from the three subtests. The validation question: does the CEI correlate with established cognitive and academic measures at levels comparable to other crystallized-IQ instruments?
Method
Respondents who completed the JCCES were drawn from a multi-year online sample. Subgroups within this sample had also completed, or had self-reported scores from, one or more of the following measures: Reynolds Intellectual Assessment Scale (RIAS), Wechsler Adult Intelligence Scale—Third Edition (WAIS-III), Wechsler Intelligence Scale for Children—Third Edition (WISC-III), General Ability Measure for Adults (GAMA), Stanford-Binet Intelligence Scale (SBIS), Scholastic Assessment Test (SAT, three versions: pre-1995, 1995–2005, post-2005), American College Test (ACT), Graduate Record Examination (GRE), and Armed Forces Qualification Test (AFQT).
Pearson correlations were computed between the JCCES CEI and each measure (and major subscales where available). Sample sizes varied across the comparison measures, reflecting how many respondents had supplied scores on each. The analysis is convergent-validity oriented: each correlation tests whether the JCCES CEI shares the expected substantial variance with the established benchmark.
Results
Across all comparisons, the JCCES CEI correlated substantially with established cognitive and academic measures. Most correlations were significant at p < .001.
Cognitive ability batteries
- RIAS (N = 138): CEI–Verbal Intelligence Index r = .859; CEI–Verbal Reasoning r = .859; CEI–Information (Guess What?) r = .814.
- WAIS-III (N = 76): CEI–Full Scale IQ r = .821; CEI–Verbal IQ r = .837; CEI–Performance IQ r = .660; CEI–Verbal Comprehension Index r = .816; CEI–Vocabulary r = .775; CEI–Information r = .769; CEI–Similarities r = .579.
- WISC-III (N = 29): CEI–Full Scale IQ r = .851; CEI–Verbal IQ r = .665; CEI–Performance IQ r = .703.
- GAMA (N = 64): CEI–GAMA IQ r = .617; subscale loadings r = .455 to .612.
- SBIS (N = 10): CEI–Full Scale IQ r = .883 (small sample; widest confidence interval).
Academic-achievement measures
- SAT (across three test versions): CEI–Composite r = .814 (pre-1995, N = 87), r = .826 (1995–2005, N = 118), r = .858 (post-2005, N = 125). Strong relationships across all SAT versions, suggesting that the SAT–crystallized-IQ link is robust to the test’s revisions over the period studied.
- ACT (N = 133): CEI–Composite r = .691; subscale correlations r = .600 (Mathematics) to .685 (Science).
- GRE (N = 66): CEI–Composite r = .844; CEI–Verbal r = .768; CEI–Quantitative r = .819; CEI–Analytical r = .430 (N = 29; weaker, see discussion).
- AFQT (N = 62): CEI–AFQT (deviation IQ) r = .825.
What the pattern means
The JCCES CEI’s correlation with WAIS-III FSIQ (r = .821) is in the range typically observed between two well-validated IQ instruments measuring the same general construct (r values typically .80 to .90 for well-established IQ batteries; lower for narrower constructs). The CEI–WISC-III FSIQ correlation (r = .851) extends this convergent validity into the child age range. The strong CEI–RIAS Verbal Intelligence Index correlation (r = .859) and CEI–RIAS Verbal Reasoning correlation (r = .859) confirm the JCCES targets verbal-side cognitive ability specifically, consistent with its crystallized construct definition.
The differential correlation with WAIS-III VIQ (r = .837) versus PIQ (r = .660) is exactly what crystallized-IQ theory predicts: crystallized abilities are heavily verbal/knowledge-loaded; nonverbal/performance abilities load more on fluid intelligence. The CEI–GAMA correlation (r = .617), which is weaker than the CEI–WAIS-III correlation, fits the same logic: the GAMA is purely nonverbal and figural, so a crystallized-IQ measure should correlate with it less strongly than with predominantly verbal IQ batteries. Factor analysis on a different sample showed that JCCES and GAMA load on separable factors, which is consistent with the moderate (but real) correlation observed here.
The academic-achievement correlations follow the same predictable pattern. SAT scores correlate strongly with general intelligence (Frey & Detterman, 2004 reported r ≈ .82 in similar samples). The CEI–SAT correlations of .814 to .858 are in the same range, supporting the case that the JCCES taps the same underlying ability that drives SAT performance. The CEI–ACT correlations are slightly weaker (.691 composite) but still substantial, with subscale variation reflecting that ACT subscales tap somewhat different cognitive demands. The CEI–GRE correlation of .844 confirms the pattern at the graduate-admissions level.
The GRE Analytical anomaly
The one anomaly in the otherwise consistent pattern is the CEI–GRE Analytical correlation of r = .430 (N = 29). Two interpretations are plausible. First, the Analytical subscale of the GRE measures analytical writing — articulating complex ideas, supporting arguments with reasons and examples — which is a different cognitive demand from the JCCES’s verbal-analogy / mathematical-problem / general-knowledge content. The moderate correlation likely reflects that the JCCES does not directly tap analytical-writing performance, even when it does tap general crystallized intelligence. Second, the small subsample size (N = 29) produces a wide confidence interval around the point estimate; the true correlation could plausibly fall anywhere from about .20 to .65.
Both interpretations are compatible with the broader pattern: the JCCES CEI is a strong measure of general crystallized intelligence and broad academic achievement, but it does not capture every specific subskill that targeted assessments measure. The GRE Analytical is the cleanest example of a subskill that the JCCES does not directly index.
Methodological caveats
Sample sizes vary substantially across measures, from N = 10 (SBIS) to N = 138 (RIAS). The smaller samples produce wider confidence intervals around the correlation estimates, and the most interesting reading of those numbers is qualitative (“the JCCES correlates strongly with the SBIS in this small sample”) rather than quantitative (“the population correlation is .883”). Replication in larger samples for the smaller-N comparisons would tighten the inference.
Selection bias is a recurring concern in voluntary online samples. Respondents who complete a cognitive battery online and supply scores from other instruments may differ systematically from the broader population — typically toward higher cognitive ability, higher educational attainment, and greater interest in psychometric self-assessment. The correlations reported here may overestimate the true correlations in less selected populations, though the direction and substantial magnitude of the relationships should generalize.
Self-reported scores (used for SAT, ACT, GRE, and some other measures) are subject to reporting biases, including social-desirability effects and recall errors. The JCCES itself was administered directly (not self-reported), but the comparison-instrument scores depend on respondents’ memory and willingness to disclose. This is a constraint on causal interpretation but does not undermine the convergent-validity argument: the pattern of correlations across many measures and many self-report-vs-direct combinations would be difficult to produce if the JCCES were not measuring something close to general crystallized intelligence.
Implications for practice
For the JCCES specifically: the convergent-validity case is strong. The CEI correlates with WAIS-III FSIQ at the level expected for two valid IQ measures, with the SAT at the level expected for a crystallized-IQ measure correlated with a heavily-verbal academic test, and with the GRE composite at the level expected for a crystallized-IQ measure correlated with a graduate-admissions test. The pattern of differential correlations — stronger with verbal/crystallized measures, weaker with nonverbal/fluid measures — confirms the construct specificity.
For users considering the JCCES in research or applied contexts: the test functions as a reasonable proxy for general cognitive ability, particularly its crystallized side. It does not replace full-battery IQ assessment for clinical purposes, but it provides a fast, online-administered measure that correlates with full-battery results at validation-ready levels. For research applications where cognitive ability is a covariate or moderator rather than the focal outcome, the JCCES CEI is a defensible choice with documented convergent validity.
For test developers more generally: the validation pattern documented here is the standard against which new cognitive tests should be evaluated. A test that correlates with established benchmarks at r > .80 across multiple comparison measures has empirical convergent validity; lower correlations across the board suggest a measurement problem; high correlations on some measures and low on others suggest construct specificity (which can be a feature, depending on what the test claims to measure).
Frequently Asked Questions
What does it mean that the JCCES CEI correlates with WAIS-III at r = .821?
It means the two tests share about 67% of their variance (r² = .67), which is the typical level of overlap between two well-validated IQ measures of the same general construct. Higher would suggest near-perfect redundancy; substantially lower would suggest the JCCES is measuring something different from established IQ.
Why is the CEI–GRE Analytical correlation weaker than the others?
The GRE Analytical subscale measures analytical writing — articulating complex ideas and supporting arguments — which is a different cognitive demand from the JCCES’s verbal-analogy, mathematical-problem, and general-knowledge content. The moderate correlation (r = .430) reflects that the JCCES does not directly tap this specific subskill. The small N (29) for this comparison also widens the confidence interval substantially.
Are these correlations in the typical range for crystallized-IQ measures?
Yes. Established crystallized-IQ measures correlate with each other at r values typically in the .75–.90 range; the JCCES CEI’s correlations with WAIS-III VIQ (.837), RIAS Verbal Intelligence Index (.859), and SAT (.814–.858) all fall in this range. The pattern of stronger correlations with verbal-loaded measures and weaker with nonverbal measures is also typical.
Can the JCCES replace a full IQ battery?
For research applications where cognitive ability is a covariate or moderator, yes. For clinical applications requiring a full cognitive profile (clarification of strengths and weaknesses across multiple cognitive domains, support for diagnostic decisions, eligibility determinations), no — full-battery instruments like the WAIS or WISC are designed to provide the multi-domain coverage that a three-subtest crystallized-IQ measure cannot match.
How was the multi-instrument sample assembled?
Respondents who completed the JCCES were asked to supply scores or take additional tests from established cognitive and academic measures. Sample sizes varied across the comparison measures (N = 10 to 138) reflecting how many respondents had supplied scores on each; the larger samples are correspondingly more reliable estimates.
References
- Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analytic studies. Cambridge University Press. https://doi.org/10.1017/CBO9780511571312
- Deary, I. J., Strand, S., Smith, P., & Fernandes, C. (2007). Intelligence and educational achievement. Intelligence, 35(1), 13–21. https://doi.org/10.1016/j.intell.2006.02.001
- Horn, J. L., & Cattell, R. B. (1966). Refinement and test of the theory of fluid and crystallized general intelligences. Journal of Educational Psychology, 57(5), 253–270. https://doi.org/10.1037/h0023816
- Neisser, U., Boodoo, G., Bouchard, T. J., Boykin, A. W., Brody, N., Ceci, S. J., Halpern, D. F., Loehlin, J. C., Perloff, R., Sternberg, R. J., & Urbina, S. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51(2), 77–101. https://doi.org/10.1037/0003-066X.51.2.77