The shape of the distribution
IQ scores are constructed to follow a normal distribution with a mean of 100 and a standard deviation of 15. That shape is imposed by the scoring, not discovered in nature: raw scores are transformed so the population fits the bell curve.
The consequence is that scores cluster tightly in the middle and thin out fast at the edges. Roughly 68% of people fall between 85 and 115, about 95% between 70 and 130, and fewer than one in four hundred above 145.
Band by band
Below 70 (about 2% of people): historically used as one criterion for intellectual disability, but a diagnosis requires assessment of adaptive functioning as well and can never rest on a test score alone.
70–84 (about 14%): below average on the test. Many people in this band work, study, and live entirely ordinary lives; the score describes performance on abstract reasoning tasks, not competence.
85–114 (about 68%): the average range. A 92 and a 108 are far closer together in practice than the numbers feel, and the gap between them is comparable to normal day-to-day variation in a single person.
115–129 (about 14%): above average. This is roughly the range where abstract, information-dense work becomes comfortable rather than effortful.
130 and above (about 2%): the range often labelled 'gifted' and used as a cutoff by high-IQ societies. Precision is poor here because there are few items difficult enough to separate people cleanly.
Percentiles are easier to reason about
Converting to percentile removes the false precision of the raw number. 100 is the 50th percentile, 110 is roughly the 75th, 120 the 91st, 130 the 98th, 140 the 99.6th.
Notice how quickly the percentile saturates. The distance from 100 to 115 covers about a third of the population; the distance from 130 to 145 covers under 2%. Comparing high scores to each other is comparing points on a very flat part of the curve.
Why labels do harm
Twentieth-century test manuals attached clinical labels to low bands and words like 'superior' and 'very superior' to high ones. Several of those terms became slurs; all of them encourage reading a percentile as an identity.
Modern practice prefers descriptive statements: 'performed better than about 84% of the reference sample on non-verbal reasoning'. It is longer, and it is the only version that is actually true.
The error bar you never see
A reported score is an estimate with measurement error. Clinical batteries usually publish a 95% confidence interval of about plus or minus 5 points; short unsupervised tests deserve a wider band, often ten points or more.
This is why chasing a two-point difference between two attempts is meaningless, and why a single number should never be the sole basis for any decision about schooling, work, or self-worth.
Sub-scores matter more than the total
Two people scoring 110 can have entirely different profiles: one strong in spatial rotation and weak in working memory, the other the reverse. The composite hides exactly the information that is useful for study strategies or task design.
When you read any report, look for the breakdown and the largest gap between domains. A pronounced imbalance is often more informative than the headline figure, and in clinical settings it is what prompts closer investigation.