Where the Genetic Overlap of Autism and Schizophrenia Stops

What the research found

For most of the twentieth century the two diagnoses were understood to be relatives, joined by a common cause. Frieda Fromm-Reichmann gave that cause a name in schizophrenia in 1948, the “schizophrenogenic mother.” Leo Kanner, describing eleven children with autism in 1943, recorded their parents as intelligent, of moderately high social standing, and unwarm toward their children, and Bruno Bettelheim’s phrase “refrigerator mother” carried the idea into psychoanalysis. Two conditions, one explanation, and the explanation was the same woman.

Twin research displaced that account. Michael Rutter argued in the 1970s that autism was a congenital cognitive disorder under strong genetic influence, and adoption and high-risk studies established the same for schizophrenia. A 2020 review in Journal of Neural Transmission by Akira Imamura and colleagues at Nagasaki University assembles what the method has produced for both conditions. It is a narrative review, with no stated search protocol and no pooled statistic of its own.

The design it summarises is old and simple. Identical twins share essentially all of their genetic information and fraternal twins about half, so a diagnosis that appears in both members of a pair far more often when the pair is identical is a diagnosis under genetic influence. Danish national registers supply the largest schizophrenia estimate in the review: Hilker and colleagues examined more than 30,000 twin pairs and found the diagnosis in both members of 33 percent of identical pairs and 7 percent of fraternal pairs, a heritability of 79 percent. Older and smaller studies reported identical-pair rates closer to 48 percent.

Autism’s estimates disagree with each other. Sandin and colleagues, working with 37,570 twin pairs alongside more than two and a half million full sibling pairs, put heritability at 83 percent, with a confidence interval of 79 to 87. Hallmayer and colleagues, working with 192 twin pairs, put it at 37 percent for autism and 38 percent for autism spectrum disorder, with intervals wide enough to be close to uninformative, and attributed a larger share to environment shared by the pair. The review reports both figures without reconciling them.

Genetic overlap between psychiatric conditions enters the review late and briefly. Genome-wide work has identified risk loci with shared effects across five major psychiatric disorders, and a twin study by Taylor and colleagues of roughly 5,000 UK pairs compared autistic traits with psychotic experiences in adolescence across an unselected community sample. Autistic traits were weakly correlated with paranoia and hallucinations and modestly correlated with cognitive disorganization. The same review reports a strong correlation between autistic traits and ADHD traits.

What it means for you and your child

In the Danish registers, where one identical twin held a schizophrenia diagnosis, the other did not in roughly two thirds of pairs, despite sharing essentially all of their genetic information and much of their upbringing. Heritability is calculated in this literature from the gap between identical and fraternal concordance across a whole population. It describes how much of the variation in a population tracks genetic variation, and it forecasts nothing about an individual child.

Shared genetic influence between psychiatric conditions is the common case. The review lists it across schizophrenia, autism, ADHD, bipolar disorder and depression, and notes proposed overlap with disorders of the immune system as well. Two diagnoses drawing on some of the same genetic variation is evidence that the categories are imprecise at their edges, and it describes no route by which one condition becomes the other.

Because the twin studies were built to partition variance across populations, the review supplies no figure for how often an autistic child later receives a schizophrenia diagnosis. That number is absent, and the correlations it does report describe how two sets of trait scores moved together across thousands of adolescents in the general population. The autism trait scores in that study belonged largely to people who had no diagnosis of anything.

One mechanism appears in the review, and it runs in opposite directions for the two conditions. Synaptic pruning, the developmental removal of unused connections between neurons, has been reported as excessive in schizophrenia and as poorly targeted in autism, occasionally leaving an excess of connections. The review presents this as a hypothesis under investigation.

Uncertainty is the review’s own conclusion. Its authors state that the molecular twin work has not reached a consensus, and the heritability estimates for autism still run from 37 percent to 83 percent depending on which sample is counted. The history in this field is one of confident explanations later withdrawn, and the current evidence connecting these two conditions is weak and was measured in traits. Anyone offering a family a firmer version of that connection has gone beyond what the traits were able to show.