30 Oct The Blueprint Metaphor
I have more to say about Plomin’s use of my ideas in Blueprint, but for now I want to start at the beginning and write about the book and the ideas in contains. The obvious place to begin is the title. It is safe to say that there is not a biologist on the planet who likes the blueprint metaphor. Here is an excellent twitter thread from Simon Fisher:
Your genome is not a blueprint. A thread about misleading metaphors in science communication… 1/11— Simon E. Fisher (@ProfSimonFisher) July 15, 2018
Paige Harden also discusses the issue in her review of Blueprint:
I actually felt some optimism about the book when I first saw the title. I like controversial ideas and the analysis that supports them. Maybe Plomin was going to make the bold claim that we have misunderstood the blueprint metaphor, that DNA is actually more like a blueprint than we realize, or that the relationship between blueprints and finished products is different that what we usually think. But no. Setting the tone for the entire book, Plomin just puts the blueprint metaphor on the table, pounds it a few times for emphasis, and moves on. Here, in its entirety, is his defense of the metaphor:
As I will demonstrate in this book, the DNA differences inherited from our parents at the moment of conception are the consistent, lifelong source of psychological individuality, the blueprint that makes us who we are. A blueprint is a plan. It is obviously not the same as the finished three‑ dimensional structure – we don’t look like a double helix. DNA isn’t all that matters but it matters more than everything else put together in terms of the stable psychological traits that <make us who we are.>
The obvious problem with the blueprint metaphor is that the relationship between blueprints and buildings is nothing like the relationship between genotypes and organisms. A blueprint of a house has a roof and walls and windows and a floor that have a direct one-to-one correspondence with the roof, walls, windows and floor in the house. There is nothing like this in DNA: there aren’t head genes and feet genes, much less extraversion genes and divorce genes.
Presumably Plomin knows this. I have seen some defenses of the Blueprint metaphor that said, come on, it’s a popular book, it’s just the title, just a metaphor. He needs to get people’s attention. But the problem is, when you examine the blueprint metaphor more closely you see that it is closely tied to the deeper message of the book: that genetic variation is more than just a crucial part of the intricate biological machinery that creates organisms, it is what determines what individual people become.
So the thing about blueprints is not just that they have a one-to-one correspondence to buildings that DNA doesn’t have. A blueprint of a house has more than just a probabilistic relation to the house that is eventually built. We wouldn’t say, if a particular building starts out with a blueprint of a house, it increases the probability that the eventual constructed building is also going to be a house, as opposed to a church or an office building or a giraffe. A blueprint represents a specific thing, a single outcome. We can use blueprints as a causal explanation. Why does that building have a flat roof? Because that’s the way it was drawn in the blueprint. Blueprints with flat roofs create buildings with flat roofs.
I am about to suggest my alternative to the blueprint metaphor, but first I need to offer some excuses. Blueprint discussions sometimes proceed as though there is a scientific rule that there must exist some kitchen- or construction-based metaphor that exactly captures the way DNA works. There isn’t. DNA works like DNA, and any simple metaphor for it is going to be inaccurate in lots of ways. That’s why I might not have objected to the blueprint metaphor if Plomin had taken the time to actually think about it. The problem with blueprint is not so much that it is inaccurate, like any metaphor, it is that it is systematically biased in a way that allows Plomin to sell a dangerously deterministic conclusion. I’ll have more to say about that.
Anyway, here goes. DNA is like a big pile of raw materials that are used to create a building. Under the right circumstances, DNA assembles itself (there is no builder) into an organism, along the way interacting with the environment and taking in much more stuff from the outside. “Raw materials” may be the wrong phrase– as Plomin says, organisms aren’t built out of DNA. What I mean is that DNA is the starting place of an unimaginably complex developmental process that eventually self-organizes into a working organism.
Simon Fisher illustrated this with a wonderful figure from @WiringTheBrain:
…Why isn't reverse engineering feasible? Because genomes guide the building of bodies through networks of gene activity interacting with other intrinsic & external factors via developmental cascades. This image, courtesy of @WiringTheBrain, elegantly captures the point…8/11 pic.twitter.com/DhjsCulXV5— Simon E. Fisher (@ProfSimonFisher) July 15, 2018
Anyway, I’m sure people can come up with a lot of problems with that metaphor, but I want to make a particular point. Even though the pile of materials has none of the blueprint qualities– there is no section of the pile that corresponds to the roof, and no way to identify from looking at the pile exactly what the eventual building is going to look like, two interesting things continue to be true.
First, it turns out that if you have a way of quantifying how similar any two piles are– in general, do they contain the same kind of materials, are the piles roughly the same shape, etc– that similarity in starting place winds up being correlated with how similar the eventual buildings are, for pretty much any building characteristic you can come up with. So pile-similarity is correlated with similarity in how the buildings are used, or what color they are or how big they are, or whatever. These correlations aren’t enormous, but they are striking, often in the range of .4-.6. What’s more, it turns out that occasionally, there are identical piles of materials, and although these identical piles don’t produce identical buildings, the buildings they produce are damn similar, often in the range of .7-.9. This is the heritability of building type.
Somewhat later, it was discovered that there was another way to see the relationship between the piles and the buildings. If you measure a whole lot of particular characteristics of the piles– especially how much of ingredient X and Y and Z they contain– there turns out to be tiny but detectable correlations between these measurements and the eventual buildings. The amount of ingredient X is correlated .007 with the probability that the eventual building will be a church. In enormous samples of piles, some of the tiny correlations are significant. What’s more, it turns out that if you add up all of the little ingredient measures into a single score, you get something that has a higher correlation, maybe .2 or so. This is GWAS and PGS.
You get the idea. Here is the point– although building heritability and building GWAS are both potentially interesting and might have some useful scientific applications, neither one of them implies that there is any deterministic relationship between pile characteristics and building characteristics. So the fact that piles that contain more ingredient X are slightly more likely than other piles to eventually become churches does not mean that ingredient X is an essential component of churches. The right way to say it would be, right now, under current circumstances and building practices, piles with ingredient X are more likely to become churches. Of course, building practices change, ideas about churches change, and tomorrow a different set of pile characteristic might be associated with churches. Of course, a scientist might believe that a particular ingredient is an essential component of churches, but proving that would require additional evidence, presumably in the form of a well-specified construction (ie, biological) mechanism that shows how ingredient X deterministically leads to churches under most plausible circumstances.
OK, enough for today. “Blueprint” is trying to sell the idea that a bunch of mid-range ingredient correlations can be the basis for a deterministic developmental biology and psychology, but they can’t. In what may well be the worst sentence ever written about behavior genetics, Plomin says, “Put crudely, nice parents have nice children because they are all nice genetically.” The “put crudely” qualification is no excuse. It is the equivalent of saying that churches are churches because they started out with churchy materials.
[Edited for typos 10/30/18]