Free IQ Test: 32 Questions, Honest Scoring

The test samples four kinds of reasoning that psychologists use to estimate general cognitive ability, often called g. They're the same four item types used in the International Cognitive Ability Resource (ICAR), a public-domain test developed by academic researchers and validated on large online samples.

• Number and letter series (8 questions): spot the rule behind a sequence and extend it.
• Verbal reasoning (9 questions): analogies, codes, and short logic puzzles. They're built to test reasoning, not vocabulary.
• Matrix reasoning (9 questions): 3×3 grids of figures that follow one to four rules at once. This is the classic nonverbal format.
• Spatial reasoning (6 questions): fold a flat pattern into a cube in your head and pick the cube it makes.

Every question here is original. The matrix, series, and cube questions are generated fresh for each person from rule templates, so your test differs from your friend's and answers can't be passed around. Each template keeps the same rules and difficulty no matter which shapes or numbers it draws. That's what lets the scoring treat every version of a question as the same item.

Most online IQ tests count your right answers and look the total up in a made-up table. This one uses item response theory (IRT), the model behind modern standardized tests.

1. Every question has its own difficulty

Each question has a difficulty and a discrimination (how sharply it separates weaker from stronger reasoners). Scoring also accounts for lucky guesses: one in eight on a matrix question, one in five on a verbal question. Getting a hard matrix right tells us more than getting an easy series right, and the score reflects that.

2. Your score is an estimate with error bars

Your answer pattern is combined into an expected a posteriori (EAP) estimate of your ability, along with how uncertain that estimate is. That's why your result is shown as a range, not one falsely precise number. With 32 questions, a typical 90% range spans about 15 to 25 IQ points. Any honest test this short will have ranges that wide.

3. The test calibrates itself

When you finish, your answers (right or wrong for each question, plus timing) are logged anonymously. Each week a script on our server re-estimates every question's difficulty from everyone's responses, using the standard marginal maximum likelihood method (Bock & Aitkin). It leaves out rushed or repeat attempts and publishes the new values, which this page loads automatically. Until enough data exists, the test uses starting values set by design and labels results provisional.

4. Linking to the general population

People who take online IQ tests aren't a random sample of the population, so comparing you only with other test-takers would skew the scale. To fix this, the test optionally asks for a past SAT score. The College Board publishes nationally representative percentiles for SAT totals, meaning percentiles for all U.S. 11th and 12th graders, not just those who took the test. SAT scores also track general ability closely: one large study found a correlation of .82 with a measure of g. Matching scores on this test to those percentiles places the scale on a population footing. The score is labeled linked to U.S. norms only once enough SAT reports have come in.

Live calibration status

Scoring status

Loading…

These numbers update automatically when the calibration job runs. Reliability above 0.80 is good for a test this short.

More accurate than most, less accurate than a professional assessment. Here's what limits any test taken online, this one included:

• Length. Clinical IQ batteries take an hour or more across many subtests. Thirty-two questions can rank people well but can't pin down a score to a few points.
• Conditions. Nobody is supervising. Distractions, a small phone screen, being tired, or looking things up all shift the result.
• Norms. A professional test is normed on a sample chosen to match the population by age, education, and region. Here, norms come from statistical calibration and self-reported SAT scores. That's a sound method, but people tend to remember their SAT scores a bit higher than they were.
• Ceiling and floor. With 32 questions the test measures best between roughly 70 and 135. Above and below that, the range widens, and scores past 145 are shown as “145+”.
• Retakes. Scores go up the second time because the formats are familiar. Your first attempt is the one that counts.

Use your result as a reasonable estimate of where your reasoning sits, not as a diagnosis or a credential.

IQ is a standardized scale. On major tests such as the Wechsler scales, the average is set to 100, and 15 points is one standard deviation. In a normal distribution, that puts the following share of people in each band:

A score of 100 means average, not “passing.” Moving from 100 to 115 takes you from the 50th percentile to about the 84th.

A licensed psychologist gives tests like the WAIS or Stanford-Binet one-on-one, in a controlled setting, with norms by age. Those results are what schools, courts, and clinicians accept. If you need an IQ score for a gifted program, an accommodation, a disability evaluation, or a high-IQ society, this test won't qualify. Those require an approved, supervised assessment.

What this test does well is give you a free, honest snapshot. You can see your range, where your relative strengths are, and how the scoring works.

The same kinds of reasoning appear on tests that actually decide things, and those can be practiced:

• Series and quantitative patterns: SAT, GRE Quant, ASVAB
• Verbal logic and deduction: LSAT Logical Reasoning, GMAT
• Figure matrices and paper folding: CogAT, SSAT, ISEE

Unlike a raw IQ estimate, scores on those exams respond well to focused practice. Browse all free practice tests.