Discourse that caught my attention this week: January 29, 2024.
On heritability, what it means, and what information it conveys about behavioral outcomes. A brief post.
As someone very much interested and immersed in developmental psychopathology, mental illness, and psychiatric nosology, the discourse around heritability estimates, particularly as they relate to schizophrenia, that emerged this week were particularly interesting from both conceptual and methodological angles. While there has been plenty of discussion concerning mental illness and schizophrenia on Substack, often generated byalong with , this latest round of discourse seemed to emerge following a set of pieces by whose Psychiatry at the Margins is a relatively new Substack that addresses all things related to psychiatry. The piece that seemed to kick start this round of discourse is below.
This was then followed-up by this piece:
And a note from Dr. Steven Hyman posted by Atfab:
In linking to a few of these pieces,then posted the below in which he makes the case it is fair to describe schizophrenia as probably mostly genetic.
At the heart of all this discourse are the concepts of statistical ‘causality’ and conceptual ‘biological determinism.’ More directly, at the most abstract level, it is a question of nature vs. nurture. The nurture here being environmental influences on schizophrenia. If schizophrenia is mostly genetic, this implies it is mostly NOT a consequence of social or environmental factors that can be potentially addressed to prevent or ameliorate the condition. Thus, the specter of biological determinism looms large.
For his part, Atfab (and Hyman) caution against categorical thinking in regards to mental illness and outline the limitations that heritability estimates derived at the population-level have for addressing causality at the individual (person) level. Both critiques have merit, but abandoning the concept of heritability doesn’t follow in my view. Atfab cites Moore & Shank’s ‘the heritability fallacy’ paper which argues heritability has no meaning at the individual level. However, this paper makes a similar mistake as did Lewontin when he argued against the classification of Homo sapiens into distinct groups because much more of human variation was captured ‘within populations’ than between them.
It is true that heritability will not explain causality in the Pearl sense at the within-person level for a disease as complex as schizophrenia, even if that complexity is mostly genetically-mediated. But, a measure of heritability (e.g., in the form of a polygenic risk score; PRS), will provide a reasonably precise prediction of an illness that, while not unambiguously taxonic (i.e., categorical), approaches a taxonic illness. This is because a non-negligible amount the genetic variation between those diagnosed as schizophrenic vs. those not is captured at this between-person level. This does not mean that important information does not exist at the within-person level in schizophrenia, but rather that heritability estimates of schizophrenia do allow one to distinguish with some accuracy ‘schizophrenics’ from ‘non-schizophrenics’ in the larger population. This feature of heritability estimates has enormous social-political implications, ranging from involuntary commitment to criminal culpability, that often fall along ideological lines. The circumstances surrounding the death of Jordan Neely on the New York City subway highlights these issues with remarkable poignancy.
More broadly, the question of what, precisely, heritability estimates can and should motivate mental health professionals to make decisions about for various mental illness conditions must involve a nuanced understanding of both degree of disease taxonicity and the differences between causality and prediction. Breiman’s seminal paper “Statistical Modeling: The Two Cultures” comes to mind:
The Multilevel Mailer is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.