The Problem with GWAS
had a Twitter moment the other day, when I came across a Tweet from Toby Young @toadmeister, about a piece in the MIT Technology Review by Anthony Regalado.
Plomin's treatment of "The Gloomy Prospect" is not quite as egregious as his taking credit for the First Law of Behavior Genetics. On the other hand, it provides a clear insight into his overall rewriting of the history of developmental behavioral genetics and the consequences that history has for his understanding of modern genomics. Let's review the history of the idea.
The phrase, "The Gloomy Prospect" comes from Plomin and Daniels (11-23-18); I will attribute the idea to Plomin moving forward. In that classic paper, Plomin observed what I would later call the second and third laws of BG: most of the variability in the typical twin study winds up in the error term, what had come to be called the "nonshared environment." Twins and siblings are similar mostly for reasons that can be broadly characterized as genetic; being raised in the same family, the so-called shared environment, or C, often doesn't seem to make to siblings more similar at all; but even identical twins raised in the same family are quite different from each other, and something must be making them different. What could it be?
Plomin and Daniels had a straightforward answer: if it is the nonshared rather than the shared environment that is making siblings different, we should focus on our attention on the systematic ways that siblings environments differ. Siblings raised together share a socioeconomic status, but they don't share a peer group, and presumably their parents don't treat them identically. So Plomin reasoned that it must be differences in siblings environments that produce the environmental causes that make siblings different. They formulated a three step research program to study the idea:
Research on nonshared environment can be categorized into (a) analyses of the magnitude of the nonshared environment component of variance, (b) attempts to identify specific nonshared factors that are experienced differently by siblings in a family, and (c) explorations of associations between nonshared factors and behavior.
What is important about Plomin's hypothesis is that it was the systematic effects of measurable environmental differences that made siblings different. Plomin specifically contrasted that hypothesis with the gloomy prospect, which he explicitly dismissed:
One gloomy prospect is that the salient environment might be unsystematic, idiosyncratic, or serendipitous events such as accidents, illnesses, or other traumas .... Such capricious events, however, are likely to prove a dead end for research. More interesting heuristically are possible systematic sources of differences between families. (p. 8)
Plomin and Daniels hypothesis was enormously influential, spawning an entire field of research into the systematic effects of environmental differences within families. Then, in 2000, Mary Waldron and I conducted a meta-analysis of that research, and reached a surprising conclusion: the hypothesis was wrong. Although the variance component called the nonshared environment accounted for upwards of 50% of the variance in twin studies, individual measured environmental differences within families accounted for practically nothing, maybe two or three percent. How could such a powerful variance component produce such tiny causal effects? The answer is the gloomy prospect. In the Three Laws paper, published that same year, I wrote:
The gloomy prospect is true. Non-shared environmental variability predominates not because of the systematic effects of environmental events that are not shared among siblings, but rather because of the unsystematic effects of all environmental events, compounded by the equally unsystematic processes that expose us to environmental events in the first place
In Blueprint, Plomin turns to the Gloomy Prospect once again:
In 1987 I wrote about this as the ‘gloomy prospect’ – the possibility that ‘the salient environment might be unsystematic, idiosyncratic, or serendipitous events’. In other words, the key environmental influence making us who we are might be down to chance, unpredictable events. To this gloomy list, I would now add that their effects don’t last. All of this makes these events extremely difficult to study. Rather than accepting this gloomy prospect at the outset, it made more sense scientifically to look for possible systematic sources of nonshared environmental effects. However, after thirty years of searching unsuccessfully for systematic non‑ shared environmental influences, it’s time to accept the gloomy prospect. Nonshared environmental influences are unsystematic, idiosyncratic, serendipitous events without lasting effects. The systematic, stable and long‑ lasting source of who we are is DNA.
In summary, Turkheimer and Waldron were right all along. But of course the story must be told without acknowledging that I-- or anyone other than Plomin-- had anything to do with it. I will now officially stop griping about my personal role in the book. There is, however, a much more important consequence to Plomin's refusal to see the difficult theoretical issue underlying the difference between robust variance components and unsystematic causal effects: the same problems apply to the genome. In the conclusion of the Three Laws paper (this was 2000, remember) is a section titled, "Anticipating the Genome Project", I wrote:
The gloomy prospect looms larger for the genome project than is generally acknowledged. The question is not whether there are correlations to be found between individual genes and complex behavior—of course there are—but instead whether there are domains of genetic causation in which the gloomy prospect does not prevail, allowing the little bits of correlational evidence to cohere into replicable and cumulative genetic models of development.
What happened since then, of course, is the failure of the candidate gene project, leading to GWAS and eventually to the GPS that Plomin now pins all his hopes on as the "stable and long-lasting source of who we are." But it is not possible to understand that progression without having a clear-eyed view of what happened in the nonshared environment. I'll turn to that story in the next couple of posts.