When people talk about IQ they usually mean one number. Psychologists have known for a century that the number hides at least two different things. Fluid reasoning is the ability to solve a problem you have never seen before: spot the pattern in a matrix, plan moves ahead. Crystallized knowledge is what you have accumulated: vocabulary, reading, facts, the things culture and experience put into you. They rise and fall differently across life. A study published in Nature Communications in April 2026 shows they also have different genetics, and that the difference matters a great deal for autism and ADHD.
What the Researchers Did
Diego Londono-Correa, Elliot Tucker-Drob, Paige Harden, Ian Deary and colleagues (University of Texas at Austin and University of Edinburgh) ran a genome-wide analysis of crystallized knowledge in about 438,000 people. They found 78 regions of the genome associated with it, five of them never linked to any cognitive trait before.
Then they separated cognition into three components, plus one more: reaction time (press a button as fast as possible), fluid reasoning (matrices, tower planning, trail-making), crystallized knowledge (vocabulary, word reading, spelling), and the noncognitive skills behind educational attainment, such as persistence. For each one they computed the genetic correlation with psychiatric conditions: how far the DNA variants that push one trait up also push the other up or down.
Four Conditions, Four Different Patterns
ADHD. Negative genetic correlations with fluid reasoning, with crystallized knowledge and with the noncognitive skills of education, and a slightly faster reaction time. The same variants that raise the odds of ADHD tend, on average, to go with lower test and school performance, but quicker simple responses.
Autism. A positive genetic correlation with crystallized knowledge, no significant link with fluid reasoning or reaction time, and a negative one with the noncognitive skills of education. Variants associated with autism go, on average, with more accumulated knowledge, and with a school path that is nevertheless harder.
Schizophrenia and bipolar disorder. The most striking pattern: lower processing speed and fluid reasoning (correlations of about -0.16 to -0.28) but higher crystallized knowledge and noncognitive skills (about +0.11 to +0.15). The authors note this fits an old observation: relatives of people with these conditions are often unusually creative, curious and accomplished. We touched on it in autism, schizophrenia and cognitive evolution.
Alzheimer's disease. Linked to lower fluid reasoning only, not to crystallized knowledge.
One more result: crystallized knowledge has a genetic correlation of 0.47 with openness to experience, the personality trait of curiosity. Knowledge is partly what a curious mind collects.
Why a Single IQ Number Misleads
Add these signals into one "general intelligence" score and they cancel out. That is why older studies found almost no genetic link between autism and IQ, or a confusingly positive one: the positive link sits in knowledge, not in reasoning. It is also a genetic echo of something clinicians see every day, the spiky profile: a very strong vocabulary next to slow processing, or sharp reasoning next to poor school results.
The authors also offer an answer to an old puzzle: if these variants raise the risk of serious conditions, why has evolution not removed them? Because a variant that costs something in one domain and gives something in another can be kept in the population at a moderate frequency. Geneticists call it balancing selection.
What This Does Not Say
A genetic correlation is a statement about populations, not about you. It does not say that people with ADHD are less intelligent, or that autistic people know more. It says that, across hundreds of thousands of genomes, the variants tilt in those directions on average. Within each group the spread is enormous, and ADHD in particular lowers test performance through attention, without touching ability.
The authors list their own limits: participants are almost all of European ancestry, some residual population structure may remain in the signal, and nearby variants on a chromosome do not necessarily act through the same biology.
Measuring Your Own Profile
The practical lesson is to look at the parts rather than the total. Our cognitive tests are split the same way. Verbal IQ (analogies and vocabulary) is mostly crystallized knowledge. Visual IQ (Raven-style matrices) is the classic measure of fluid reasoning, and Spatial IQ and Numerical IQ complete the picture. Taking them separately, or together in the complete assessment, shows you where your peaks and dips are, which is far more useful than one number.
If a large gap between your scores sounds familiar, the giftedness screening, the ADHD screener and the AQ-10 are the natural next steps. An uneven profile is not a flaw in the measurement. It is often the most informative thing the measurement found.
Source: Londono-Correa D, de la Fuente J, Davies G, Cox SR, Deary IJ, Harden KP, Tucker-Drob EM. Crystallized and fluid cognitive abilities have different genetic associations with neuropsychiatric disorders. Nature Communications, 30 April 2026. DOI 10.1038/s41467-026-72477-7. Open access. Figures are from the article; the explanations in plain language are ours. Our tests are screening tools, not a clinical IQ assessment.