The most reliable everyday signals of high cognitive ability are an unusually large vocabulary, fast pattern learning in unfamiliar domains, and comfort holding several competing ideas open before deciding. Most of the traits that circulate as signs of genius, including talking to yourself, staying up late, and being messy, have either weak evidence or none. Self-assessment is the weakest method of all, because the people least able to evaluate their own reasoning are the ones who most overestimate it.
Signals with reasonable support
Vocabulary size. This is the strongest single everyday indicator, and it is why verbal subtests carry so much weight in full IQ batteries. Word knowledge accumulates through years of reading and inference from context, so it encodes learning rate and exposure at once. It correlates with full-scale IQ more highly than almost any other single measure.
Speed of learning in a new domain. Not knowledge already held, but how quickly structure gets extracted from something unfamiliar. Picking up the rules of a new game from watching two rounds is closer to what IQ tests measure than knowing a lot of facts.
Working memory in conversation. Following an argument with three branching conditions without losing the thread. Working memory is one of the most heavily loaded components in cognitive testing and it is visible in ordinary conversation.
Tolerance for unresolved questions. Holding a question open rather than reaching for the first available answer. This looks like a personality trait and it correlates with measured ability, partly because premature closure is what produces most wrong answers on reasoning items.
Metacognitive accuracy. Knowing which of your own answers are shaky. People who score well tend to be better calibrated about their own uncertainty, which is a more useful skill than confidence.
Popular signs that do not hold up
Talking to yourself. Self-directed speech aids working memory on some tasks, which is the grain of truth. It is not a marker of high ability, and the studies get cited far past what they support.
Being a night owl. Some studies find a weak association between eveningness and cognitive ability. The effect is small, confounded by age and occupation, and does not survive as a diagnostic.
Messiness. This traces to a single small study on creative idea generation, which is not the same construct as intelligence. It has been overstated to the point of folklore.
Sarcasm and dark humour. Comprehending irony requires modelling another person's intent, which is cognitively demanding. Using sarcasm frequently is a personality trait rather than a measure of anything.
Overthinking or anxiety. The claimed link between worry and intelligence rests on thin evidence and it appeals because it reframes a difficulty as a hidden strength. Anxiety measurably depresses test performance rather than indicating high ability.
Why self-assessment fails
Self-estimated IQ correlates with measured IQ at around .3, which is weak. People are poor judges of their own cognitive ability in both directions, and the errors are not random.
The Dunning-Kruger pattern is the well-known half: the skills needed to perform well in a domain are often the same skills needed to recognise good performance, so the least able are the least equipped to notice. The less-discussed half is that high performers frequently underestimate themselves, because they assume tasks they find easy are easy for everyone.
There is also a sampling problem. Most people's sense of where they sit comes from comparison with the people around them, and those people are usually filtered by education, occupation, and social circle. Someone in a research group is comparing against a sample that is nothing like the general population.
This is the honest case for taking a test rather than reasoning about it. A test with any calibration behind it beats introspection.
What high intelligence does not predict
Rationality is a separate construct with its own measures. Cognitive ability and the tendency to reason well about evidence correlate only moderately, and high-ability people are perfectly capable of motivated reasoning. They are frequently better at constructing defences for conclusions they arrived at for other reasons.
Expertise requires accumulated practice, and ability speeds acquisition without substituting for it. Wisdom, in the sense of good judgement about how to live, has almost no measured relationship with IQ.
Life outcomes are predicted moderately at best. Cognitive ability predicts job performance and academic results, more strongly as complexity increases, but conscientiousness, health, and circumstance account for large shares of the variance. Treating a score as a forecast is a misreading of what it measures.
Questions people also ask
Can you tell someone's IQ from talking to them?
Not precisely, though not at zero either. Vocabulary and the ability to follow complex argument are audible in conversation and both correlate with measured ability. What conversation cannot separate is ability from education, confidence, and familiarity with the subject at hand.
Do intelligent people have fewer friends?
This comes from one study finding that people with higher cognitive ability reported lower life satisfaction from frequent socialising. It is a small effect about satisfaction, not about friendship counts, and it has been stretched into a claim the data does not support.
Keep reading
- IQ Percentile Chart: What Your Score Actually RanksThe full table from 70 to 145, and how percentiles are derived.
- The Big Five Personality Traits ExplainedWhat each of the five traits predicts, and why it beat the MBTI.
- Are Online IQ Tests Accurate?What separates a decent online test from a worthless one.