How our tests work

Every question on NineTraits is written by us. The tests run in your browser and your answers are not sent to our servers, which is good for privacy but also means we can't quietly check our scoring against thousands of results. So below we say plainly what each score is based on and how much weight it can take. Last reviewed on 27 October 2026.

The IQ test

What it is

A timed reasoning test: 30 questions drawn at random from a pool of 60 (8 patterns, 8 number series, 7 verbal, 7 logic), with 20 minutes in total. Every question has one correct answer, and you can review all of them at the end.

How the score is calculated

We count your correct answers and place that count on the IQ scale, where 100 is average and the standard deviation is 15. To do that we need to know how an average adult scores on our questions. We don't have a large, representative sample yet, so we use a provisional assumption: an average adult gets about 17 of 30 right, with a spread (standard deviation) of 5. Each extra correct answer is then worth about 3 IQ points. Scores are capped between 70 and 145, because a 30-question test can't tell extreme scores apart.

Why it is an estimate, honestly

• Not normed. Real IQ tests such as the WAIS or Stanford-Binet are calibrated on thousands of people chosen to match the population by age, sex, education and region. Ours is not. If our questions are easier or harder than we assume, every score shifts up or down.
• Short. Thirty questions give a rough reading. We show a range of about 7 points either side of your score; the true uncertainty may be larger.
• Not age-adjusted. Proper IQ scores compare you with people your age. Ours compares everyone with an assumed average adult, so it is least accurate for children and older adults. See how IQ changes with age.
• Conditions matter. Tiredness, distractions, a small phone screen, practice with similar puzzles and taking the test in a second language all move the score.
• Culture and language. The verbal questions depend on vocabulary in the language you take the test in. Pattern and number questions travel better, which is why your score by area is worth more than the single number.

What an IQ score measures at all is debated. The American Psychological Association's task force summary is a good, balanced starting point: Neisser and others, "Intelligence: knowns and unknowns" (1996). An official IQ score only comes from a standardised test given in person by a licensed psychologist. Never use our number for school placement, hiring, diagnosis or any official purpose.

The personality test

45 statements, five for each of nine traits, rated from "disagree" to "agree". Some statements are reversed (agreeing lowers the score) so that simply agreeing with everything doesn't push you to one end. Each trait is then scored from 0 to 100.

The traits draw on the best-supported models in personality research: the Big Five (sociability, imagination, warmth, discipline and calm map roughly onto extraversion, openness, agreeableness, conscientiousness and emotional stability) and the honesty-humility factor of the HEXACO model, plus drive, boldness and optimism. Free, public research versions of these scales are published by the International Personality Item Pool. Our wording is our own and has not been validated in a published study.

The four-letter type comes from four of the traits, each split at 50. It is a convenient label, in the spirit of Jung's types, but people don't really fall into 16 boxes: most of us score near the middle on some traits, and a few points either way can flip a letter on a retake. Type questionnaires in general are known to give many people a different type when they retake them weeks later, while trait scores are much more stable. So read your trait scores first and your type second. This test is not the Myers-Briggs Type Indicator®.

The relationship test

36 statements across nine areas that relationship research links to how happy and stable couples are, such as attachment worry and avoidance, how you handle conflict and repair, affection and closeness, plus eleven "big decisions" (children, money, where to live and so on). The areas follow well-studied research themes, such as the conflict patterns described by the Gottman Institute.

The partner scores combine how you each answered: similar values and compatible conflict styles raise "staying power", shared energy and affection raise "spark". The weights are our judgement, not a fitted model. Research suggests that personality similarity predicts relationship satisfaction only weakly (Dyrenforth and others, 2010), and that how partners feel about the relationship itself predicts more than who they are (Joel and others, 2020). A percentage cannot tell you whether a relationship will work.

The career test

36 activities, six for each of the six interest types in John Holland's RIASEC model, which is public domain and used by the US Department of Labor's O*NET career database. Your top two or three types form your code, and careers are matched by their own two-letter code. Interests are only part of a good career choice: the test knows nothing about your skills, qualifications, pay needs or local job market.

Corrections

If you think a question has a wrong or ambiguous answer, or a claim here is out of date, tell us and we'll check it. See also the disclaimer.