Method and limits
How 30 answers become a number, why that number needs an interval, and what the test expressly cannot do. Anyone who does not disclose this is selling a number with nothing behind it.
What is measured
The test estimates how well someone recognises and applies rules in unfamiliar material. It measures this in five areas with six questions each: logical reasoning on figural matrices, numerical reasoning on number series, verbal reasoning on relations of meaning, spatial reasoning on folds and rotations, and deductive reasoning on conditional statements.
What it does not measure: knowledge, creativity, persistence, social skill, motivation, diligence. These are qualities that matter in life at least as much as what is measured here.
Where the questions come from
Every question is produced by a program, following explicitly stated rules. No item is taken from a copyrighted test instrument. That has two consequences. First, we can build as many parallel forms as we like, so a second run shows different questions. Second, we know the difficulty of each question from its construction: how many rules act at once, how large the search space is, how close the distractors sit to the solution.
The distractors matter more than the solution. A question whose wrong answers are obviously wrong measures nothing. That is why every question has at least one distractor built to satisfy all the rules but one.
How the calculation works
Scoring uses a two-parameter model from item response theory with a guessing correction. Every question has two parameters: a difficulty and a discrimination. The probability of solving a question depends on the distance between your ability and the difficulty of the question.
Because with six answer options guessing sometimes succeeds, a lower guessing asymptote is added. Without it the test would systematically overestimate every value. The estimate takes the most probable ability given a standard normal distribution as prior information. That keeps a perfect or an empty result from running off to infinity.
The estimated value is then mapped onto the usual scale: mean 100, standard deviation 15.
Why every value needs an interval
The estimate yields not just a value but also how precise that value is. The mean standard error is 6.4 IQ points. From that follows a 95 per cent confidence interval roughly 25 points wide. Someone scoring 112 has a true value that most probably lies between 100 and 124.
These figures come from a measurement, not from a guess. Four thousand simulated people of known ability went through the test, and the result was a correlation of 0.898 between true and estimated value, a bias of 0.5 points, and a 95 per cent interval that contains the true value in 94.3 per cent of cases.
That is why we report every value with its interval and why we round the per-area values more coarsely than the overall value: with six questions per area, anything finer would be feigned precision.
The reference distribution, and what is provisional about it
The placement rests on the model and on a growing sample of the tests submitted here. It does not rest on a representative survey of the Swiss population. People who test here are not a random selection: those who voluntarily look for an IQ test on the internet are on average more interested, younger and more formally educated than the average.
We do not correct that away artificially, we write it down. As soon as enough data is available, the difficulty parameters of the questions will be recalibrated from the data. Until then the placement is a good approximation, not a norming in the psychometric sense.
The limits, stated plainly
This result is an orientation value from a short online test. It is not a clinical assessment and it is no basis for a school placement, a hiring decision or the evaluation of giftedness.
A value can move by several points for many reasons:
- daily form, sleep, caffeine, time of day
- interruptions and distraction during the test
- familiarity with this kind of question, that is, practice
- screen size and the legibility of the figures
- language skills, especially in the verbal part
- time pressure, which affects some people more than others
If you need a binding evaluation, see a qualified professional. In Switzerland that means the cantonal school psychology services for children and adolescents, and FSP specialist psychologists for adults.