Thanks for this comprehensive and sobering piece. I came to it by way of reading your comment on another Substack, and I’m lucky I did. I feel sheepish reading it—not because I worked in behavioural genetics, but because I was close enough to possibly (?) know better and didn’t ask myself critical questions.
I still remember my first Society for Neuroscience meeting in 2010. A leading autism researcher stood on stage, gestured at a pie chart showing the vast majority of cases labeled “idiopathic,” and said, “We’ll fill the genetics in within five years.” I was a little astounded by the chutzpah—but everyone else seemed to believe him, so I did too. Bigger samples, better tools—it felt inevitable that the field would get there, even as GWAS after GWAS failed to replicate.
But I never questioned the core assumptions: How were these traits defined? What does heritability really capture? Were any of these models falsifiable, or just endlessly adjustable? In hindsight, it was faith in the shape of empiricism.
I see the same pattern in Alzheimer’s research, where I’ve been writing about the slow collapse of the amyloid cascade hypothesis. But you could see all the signs back in the heady days of developing the first transgenic mouse model—it was all there in the very first paper that launched a thousand clinical trials. Critics have called it “too big to fail,” and that feels about right. Despite decades of failed predictions, billions in sunk cost, and no meaningful therapeutic breakthroughs, the field hasn’t let go—the originators of the hypothesis just doubled down.
And this isn’t just about behavioural genetics or amyloid. It’s about a broader problem in science: where incentives reward positive results, flashy headlines, and publication over rigor, replication, or humility. What happened to Platt’s strong inference: clear, testable hypotheses; competing explanations; and a willingness to walk away when the evidence says we should?
Thanks for this comprehensive and sobering piece. I came to it by way of reading your comment on another Substack, and I’m lucky I did. I feel sheepish reading it—not because I worked in behavioural genetics, but because I was close enough to possibly (?) know better and didn’t ask myself critical questions.
I still remember my first Society for Neuroscience meeting in 2010. A leading autism researcher stood on stage, gestured at a pie chart showing the vast majority of cases labeled “idiopathic,” and said, “We’ll fill the genetics in within five years.” I was a little astounded by the chutzpah—but everyone else seemed to believe him, so I did too. Bigger samples, better tools—it felt inevitable that the field would get there, even as GWAS after GWAS failed to replicate.
But I never questioned the core assumptions: How were these traits defined? What does heritability really capture? Were any of these models falsifiable, or just endlessly adjustable? In hindsight, it was faith in the shape of empiricism.
I see the same pattern in Alzheimer’s research, where I’ve been writing about the slow collapse of the amyloid cascade hypothesis. But you could see all the signs back in the heady days of developing the first transgenic mouse model—it was all there in the very first paper that launched a thousand clinical trials. Critics have called it “too big to fail,” and that feels about right. Despite decades of failed predictions, billions in sunk cost, and no meaningful therapeutic breakthroughs, the field hasn’t let go—the originators of the hypothesis just doubled down.
And this isn’t just about behavioural genetics or amyloid. It’s about a broader problem in science: where incentives reward positive results, flashy headlines, and publication over rigor, replication, or humility. What happened to Platt’s strong inference: clear, testable hypotheses; competing explanations; and a willingness to walk away when the evidence says we should?
Thank you for this. It’s been frustrating watching this for the past 30 years.