What Gifted Tests Actually Test: We Classified 20,000+ Practice Questions

Ask most parents what a gifted test measures and they'll say vocabulary, reading, "being smart in school." Our data says otherwise. We classified the 158,647questions in our gifted and cognitive practice banks — built battery-by-battery to the published blueprints of the CogAT, OLSAT, CCAT, WISC-V, NNAT, and Raven's — and measured what these tests are actually made of.

Key findings

• Only 26% of gifted-test content is verbal. Nonverbal reasoning — patterns, matrices, spatial logic — accounts for 49% of the combined banks, the single largest share.
• The CogAT, OLSAT, and CCAT are almost perfectly balanced thirds of verbal, quantitative, and nonverbal reasoning — meaning two-thirds of these tests don't reward reading ability at all.
• The NNAT and Raven's are 100% nonverbal — a child who has never seen a figure matrix is walking in blind, regardless of how strong a reader they are.
• WISC-V-style content is half nonverbal, with verbal and quantitative reasoning splitting the rest.

The composition, test by test

Why this matters for your child

The most common prep mistake we see is drilling vocabulary and reading — the school-shaped skills — while ignoring the format that dominates these tests: nonverbal reasoning. Pattern completion, figure matrices, and paper folding are highly learnable, but they are almost never taught in class. That's also why nonverbal-heavy tests are considered fairer for English-language learners — and why familiarity, not language, is the biggest controllable advantage on test day.

Methodology

Figures are computed live from the iPrepGenius question bank (311,000+verified, unretired practice questions at the time you loaded this page). Our banks are original practice content modeled on each test's published blueprint and format; they are not actual test items, and findings describe our bank's composition — a faithful proxy for, but not a publication of, the tests themselves. Difficulty ratings use our 1–5 scale assigned at generation and refined by calibration. Aggregated practice statistics, where shown, are fully anonymized. You're welcome to cite this study with a link.