What IQ Tests Actually Measure, and What They Leave Out
An IQ test samples a few reasoning and memory tasks, then reports a rank against a comparison group. Here is what a battery covers and what it never touches.
Topic hub · 20 articles
Almost everything written about IQ online is either a sales pitch for a quiz or an argument about politics. Neither tells you the thing worth knowing: an IQ score is a rank, not a quantity. It describes where one person sat relative to a comparison group on one day, on one set of tasks, and it carries a margin of error wide enough to matter.
That single idea does most of the work in this subject. It explains why two tests disagree, why a score moves between sittings, why "genius-level" thresholds are marketing, and why a free online quiz cannot produce the number a supervised assessment produces — no matter how similar the questions look.
These articles cover the reasoning tasks themselves, the statistics that turn answers into a score, and the limits of the whole exercise. They are written to be useful to someone deciding whether a result means anything, and honest about the fact that most of the time the answer is "less than the number implies".
Read the cornerstone guide first. Standardisation, norming and the standard error of measurement come up in every other article here, and none of them make sense in isolation.
Matrix reasoning, series completion, analogies and spatial rotation each have their own logic. Knowing how an item is built is the difference between pattern-matching and understanding.
Practising a task type improves performance on that task type. That is a real, useful effect, and it is not the same as a change in underlying ability — an important distinction the practice-focused hubs return to repeatedly.
An IQ test samples a few reasoning and memory tasks, then reports a rank against a comparison group. Here is what a battery covers and what it never touches.
Abstract reasoning items are engineered objects: a rule, a key and distractors chosen to match specific errors. Knowing the anatomy makes practice sharper.
A free online IQ test fails at the parts that make a score mean anything: controlled conditions, a real norm sample, an administrator and a retest interval.
Repeat a cognitive test and your score goes up. That is a measurement artefact, a real skill gain, or both — and telling them apart is the whole question.
One is reasoning with material you have never seen. The other is knowledge you already hold. They are measured differently and they age in opposite directions.
A shareable quiz and a normed assessment can use similar-looking questions and still be completely different objects. Here is the checklist that separates them.
An IQ score is not a count of correct answers. It is a rank inside a norm sample, rescaled so the average is 100. Here is each step of that conversion.
EQ is not a second IQ. Ability and trait models of emotional intelligence measure different things, correlate weakly, and rest on much thinner standardisation.
Score bands are cut points drawn across a continuous curve by a publisher, not natural categories. Here is how they are built and why the edges are unreliable.
Even a well-built battery, correctly administered, has hard limits: what it samples, what it cannot measure precisely, how it ages, what it should not decide.
Nonverbal reasoning tests replace words with shapes. That removes a language barrier, but it does not make a test culture-free. Here is what the evidence shows.
Numerical reasoning tests assume you can already calculate. The difficulty is deciding what to compute from a table, and doing it inside a 40-second budget.
A percentage is how much you got right. A percentile is where you ranked. A standard score is how far from the mean you sat. Three numbers, routinely confused.
Reliability asks whether a test repeats itself. Validity asks whether the conclusion drawn from the score is justified. Two questions to ask of any test.
Spatial ability and mathematics are related but distinct. What spatial reasoning means, how it is measured, and why arithmetic drill is not practice for it.
How to budget seconds per question, when to abandon an item, and what a guess is actually worth under different marking schemes — with the arithmetic shown.
Verbal reasoning tests use a short list of formats: true/false/cannot say, text completion, sentence equivalence and comprehension. Here is what each probes.
An app's brain age is a performance index mapped onto an age label by the publisher. A research brain age is derived from MRI. The two share only a name.
The average is 100 because the scale is built that way, and it is reset to 100 every time a test is renormed. That single fact reframes the whole category.
Alfred Binet and Theodore Simon built the first practical intelligence scale in 1905 to find schoolchildren who needed help. Here is how its purpose drifted.