What is a good IQ score? Good compared with whom, and for what

Good compared with whom

Good is not a property of a number. It is a comparison, and the comparison group is the norm sample of one particular test. That is what norm-referenced means: such measures inform the test-takers where they stand relative to others in the distribution. A more precise question is longer than the one people ask. Good compared with whom, on which test, measuring which abilities, and on which day. Change any one of those four and the same performance prints as a different number.

This is also why a single figure travels badly. A score describes a rank in a reference group on the abilities that were sampled. It does not describe how someone learns an unfamiliar job, recovers from a bad night's sleep, works in a second language, or performs after a fortnight of practice on the item types. Our guide to the IQ scale gives the percentile behind any particular number; what no page can give is a verdict that the number carries with it.

What cognitive test scores predict, and how much

Decades of research find that scores on general cognitive ability tests correlate positively with school grades, years of education, and performance in training and in complex jobs. The correlations are real and they are moderate: in most studies they leave the majority of the differences between people unexplained. Motivation, health, opportunity, personality, teaching and practice carry the rest. A higher score shifts the odds across a large group. It does not decide an outcome for one person, and a lower score does not close a door.

The distance between those two sentences is the whole of this page. A correlation is a statement about a population. Reading it backwards, from the population to the person in front of you, is the move that produces career lists on IQ pages, and it is the reason this site does not publish one.

When a score changes a decision, and when it does not

A score changes something when a decision formally depends on it: eligibility for an educational programme, a clinical assessment that combines the score with history and interviews, or a research study. In those settings the test is administered by a trained professional under standard conditions, and the report carries confidence intervals and subscale patterns that mean considerably more than the headline number does.

Outside those settings, a number from a browser quiz has no decision attached to it. That is not a reason to avoid reasoning tests. It is a reason to read them for what they actually show, which is the kind of problem you solved and the kind you did not.

What a number on its own does not license

A score does not license a job list, a prediction about a career, a claim about a group of people, or a label that a person carries afterwards. Those four are what score-meaning pages add most often, and none of them follows from a rank in a norm sample. The professional standards for educational and psychological testing are built on the opposite idea: a test result supports a specific interpretation for a specific use, and the evidence for that use has to be gathered rather than assumed from the size of the number.

It does not license a comparison with a figure attached to a famous name either. Those figures are retrospective estimates or inventions, no test was administered to produce them, and a comparison with them is a comparison with nothing.

How to read a MyTestAtlas result instead

Our reasoning tests show your score by question type and difficulty, list every question with your answer and the correct one, and offer a practice estimate on the IQ tests. A result such as 27 of 36, all six figure matrices solved, or two of eight hard number series missed gives you a specific place to focus your practice.