The short version
- Twin studies say ADHD is about 74% heritable [1].
- Genome-wide studies say common DNA variants account for roughly a third of that [1].
- The difference is not a contradiction. It is called the heritability gap, and researchers think part of it lies in rarer variants and structural changes that common-variant studies cannot see [1, 3] (Tier B).
The usual mix-up
"If the genes are not found, it must not be genetic." Not finding every gene is a limit of the method, not evidence against heredity. Family studies show the trait is inherited; DNA studies are still working out how.
The plain-English layer
Picture the genetic contribution as many small dials plus a few large switches.
- Small dials: common variants. Thousands, each adding a sliver of risk. Genome-wide association studies (GWAS) are built to find these.
- Large switches: rare variants, including ones that appear for the first time in a child (de novo) and missing or extra stretches of DNA. They are harder to find because few people carry any one of them. They are found by sequencing families or very large case-control groups.
| Variant type | How common | Typical effect per variant | How it is found | ADHD example |
|---|---|---|---|---|
| Common single-letter changes (SNPs) | Common | Very small | GWAS | 27 loci in 2023 [2]; 39 in a 2025 meta-analysis that added symptom scores [4] |
| Rare damaging changes, including de novo | Rare to ultra-rare | Larger | Exome or genome sequencing of families and large case-control sets | KDM5B identified as a high-confidence risk gene [3] |
| Copy-number variants (CNVs) | Rare | Can be large | Microarray or sequencing | Account for part of the heritability [1] |
The deeper layer
What the 2024 sequencing study did. It sequenced the exomes (the protein-coding parts of DNA) of 152 families, each with a child with ADHD and both parents, and found more rare, new, gene-damaging variants in ADHD cases than expected. It then combined these results with a case-control cohort of 3,206 people with ADHD and 5,002 controls. KDM5B emerged as a high-confidence risk gene, and the authors estimated that about 1,057 genes contribute to ADHD risk. The genes they found overlap with risk genes for other neuropsychiatric conditions and are enriched in pathways that the authors say suggest early neurodevelopmental underpinnings [3].
What the 2025 meta-analysis added. Using 290,134 symptom measures from 70,953 people, then combining these with diagnosis data, it found 39 independent loci (17 new). Symptoms alone gave no genome-wide significant variants, yet combining them with diagnosis increased power. The authors read this as support for clinical ADHD sitting at the extreme end of a continuous liability [4].
Why the gap may persist. The commonly proposed reasons are: rare variants and structural variants not captured by common-variant studies; the limits of twin-study assumptions; and interactions between genes and environment. These are standard explanations in the field; none is settled for ADHD specifically. (Tier B/C)
Why this matters for how ADHD is described
Findings of early neurodevelopmental pathways [3] and of a continuous liability [4] fit the classification of ADHD as a neurodevelopmental condition. The policy consequences of that are argued in the policy section of this site, not here, so that this page stays a statement of evidence.
Limits
- "About a third" and "74%" come from different kinds of study and different estimation methods. Do not subtract one from the other and call the remainder "unknown genes".
- 1,057 is an estimate of how many genes contribute, from one study. It is not a list of 1,057 confirmed ADHD genes.
- KDM5B is a single-study high-confidence finding and should be described that way until replicated.
- Nothing here implies that any one person's ADHD is explained by a rare variant.
Sources
Bibliographic details checked against PubMed.