Madole & Harden (M&H) aim to present “a clear perspective on what it does – and does not – mean for genes to be causes” (target article, sect. 1.2, para. 4). We agree that this is an important project but are unsure whether they succeed in their overall aim. The authors propose that genetic causes are like nearly all environmental causes investigated in the social sciences: They are non-uniform, non-unitary, and non-explanatory (target article, sect. 1.3, para. 1). In other words, the genetic causes pointed to by genome-wide association studies (GWASs) hold true only for specific populations, their effects are probabilistic, not deterministic, and the exact nature of their causal relevance to the production of the trait is unknown. M&H call such causes shallow causes. The mitigating effects of lithium on mania, understood as an average treatment effect from a randomized controlled trial, is an example of a shallow cause. They acknowledge that this is not the only one way in which genes play a causal role in producing traits and also identify deep genetic causes. Deep genetic causes are uniform, unitary, and explanatory. Cystic fibrosis, which is caused by two mutated copies of the cystic fibrosis transmembrane regulatory (CFTR) gene looks to provide an example of a trait with a deep genetic cause.
While this distinction promises clarity, M&H often conflate cause, genetic cause, and shallow (genetic) cause and, as a result, they fail to separate genetic causal claims from environmental causal claims presented by other social scientists. This makes causes discovered by GWASs commensurate with the environmental studies M&H hope to supplant. Second, the distinction between shallow versus deep genetic causes can be helpful, but only if applied carefully and consistently. For example, the authors make apparently contradictory claims about GWASs for educational attainment (EA): The results of GWASs “alone would not move us closer to the conclusion that genes cause educational outcomes” and “we are currently in a position to conclude that genes cause EA” (target article, sect. 3.3, paras. 2 and 3). Their distinction between shallow and deep genetic causes can help resolve this and other apparently contradictory claims about genetic causation. We understand M&H to be saying that while even the most extensive GWASs cannot support the conclusion that certain genes are deep causes of a trait, current GWASs do allow the claim that certain genes are shallow causes of a trait. This claim is not contradictory and in fact is a clear statement of their understanding of the promise of GWASs. However, M&H introduce further confusion about the nature of genetic causes.
M&H seem to say that deep genetic causes are in fact not genetic causes because they are not causes at all. For example: “Such a picture of genetic causes is entirely unwarranted when we remember what it means for something to be a cause: non-uniform, non-unitary, and non-explanatory” (target article, sect. 3.3, para. 4). Here they define a cause as non-uniform, non-unitary, and non-explanatory, implying that deep genetic causes are not just something GWASs can't detect, but in fact are not causes at all. Then, just two sentences later, M&H use the example of cystic fibrosis as a genetic cause that they say is uniform, unitary, and explanatory, that is, a deep genetic cause (target article, sect. 3.3, para. 5).
We propose that this inconsistency on the status of deep genetic causes arises from the account of causation M&H defend. As noted earlier, their account of causation is meant to cover both genetic causes and shallow genetic causes, which allows them to claim at one point that GWASs can't possibly be giving fuel to eugenicists because genetic causes by definition are non-uniform, non-unitary, and non-explanatory (target article, sect. 1.3, para. 1 & sect. 3.5, para. 1). So, what are we to make of deep genetic causes? M&H clearly want to claim that some genetic causes are shallow and some are deep, and that GWASs can only find shallow genetic causes. In fact, M&H's distinction between shallow genetic causes and deep genetic causes seems to be set up to explicitly acknowledge that there are various kinds of genetic causes. If this is right, then they need a different account of genetic cause, and possibly cause in general, in order to accommodate both shallow and deep genetic causes.
M&H briefly review a few philosophical accounts of causation, then claim to adopt a probabilistic version of Jim Woodward's (Reference Woodward2003) manipulationist account. However, it is unclear how this account fits GWASs, because there is no careful manipulation of specific variables in GWASs, a point M&H acknowledge when arguing for the superiority of within-family studies over regular GWASs (target article, sect. 3.5, para. 1). Furthermore, the distinction between shallow versus deep causes does not straightforwardly map on to much philosophical discussion of genetic causation (see e.g., Gannett, Reference Gannett1999; Lynch & Bourrat, Reference Lynch and Bourrat2017; Noble, Reference Noble2008; Northcott, Reference Northcott, Plaisance and Reydon2012; Oftedal, Reference Oftedal2005; Schaffner, Reference Schaffner2016; Sober, Reference Sober, Buchanan, Brock, Wikler and Daniels2000; Waters, Reference Waters2007). For example, M&H say thatcauses need not be mechanistic (target article, sect. 3.3, para. 11), while many philosophers, including those cited here, offer accounts ofgenetic causation as being necessarily mechanistic.
M&H's shallow versus deep distinction better tracks Eric Turkheimer's (Reference Turkheimer1998, Reference Turkheimer2016) distinction between strong and weak genetic explanations than it does distinctions in the philosophy of science literature on genetic causation. For Turkheimer a weak genetic explanation says “one way or another, genetic differences among people wind up correlated with phenotypic differences” (Turkheimer, Reference Turkheimer2016, p. 24). Neither shallow genetic causes nor weak genetic explanations determine the exact difference that certain single-nucleotide polymorphisms (SNPs) make in a phenotype but this is by design. A strong genetic explanation by contrast is “the discovery that an observed phenotypic difference is a manifestation of a specific latent genetic mechanism” (Turkheimer, Reference Turkheimer2016, p. 25). Both uncovering deep genetic causes and providing strong genetic explanations can reveal specific mechanisms.
While M&H's shallow versus deep distinction may resolve some apparent contradictions in their claims about GWASs and causation, we propose that their distinction better points to a way of separating alternate types of genetic explanations. This allows that GWASs do lead to genetic explanations of traits but only in the same sense in which non-genetic social science leads to environmental explanations of traits.
Madole & Harden (M&H) aim to present “a clear perspective on what it does – and does not – mean for genes to be causes” (target article, sect. 1.2, para. 4). We agree that this is an important project but are unsure whether they succeed in their overall aim. The authors propose that genetic causes are like nearly all environmental causes investigated in the social sciences: They are non-uniform, non-unitary, and non-explanatory (target article, sect. 1.3, para. 1). In other words, the genetic causes pointed to by genome-wide association studies (GWASs) hold true only for specific populations, their effects are probabilistic, not deterministic, and the exact nature of their causal relevance to the production of the trait is unknown. M&H call such causes shallow causes. The mitigating effects of lithium on mania, understood as an average treatment effect from a randomized controlled trial, is an example of a shallow cause. They acknowledge that this is not the only one way in which genes play a causal role in producing traits and also identify deep genetic causes. Deep genetic causes are uniform, unitary, and explanatory. Cystic fibrosis, which is caused by two mutated copies of the cystic fibrosis transmembrane regulatory (CFTR) gene looks to provide an example of a trait with a deep genetic cause.
While this distinction promises clarity, M&H often conflate cause, genetic cause, and shallow (genetic) cause and, as a result, they fail to separate genetic causal claims from environmental causal claims presented by other social scientists. This makes causes discovered by GWASs commensurate with the environmental studies M&H hope to supplant. Second, the distinction between shallow versus deep genetic causes can be helpful, but only if applied carefully and consistently. For example, the authors make apparently contradictory claims about GWASs for educational attainment (EA): The results of GWASs “alone would not move us closer to the conclusion that genes cause educational outcomes” and “we are currently in a position to conclude that genes cause EA” (target article, sect. 3.3, paras. 2 and 3). Their distinction between shallow and deep genetic causes can help resolve this and other apparently contradictory claims about genetic causation. We understand M&H to be saying that while even the most extensive GWASs cannot support the conclusion that certain genes are deep causes of a trait, current GWASs do allow the claim that certain genes are shallow causes of a trait. This claim is not contradictory and in fact is a clear statement of their understanding of the promise of GWASs. However, M&H introduce further confusion about the nature of genetic causes.
M&H seem to say that deep genetic causes are in fact not genetic causes because they are not causes at all. For example: “Such a picture of genetic causes is entirely unwarranted when we remember what it means for something to be a cause: non-uniform, non-unitary, and non-explanatory” (target article, sect. 3.3, para. 4). Here they define a cause as non-uniform, non-unitary, and non-explanatory, implying that deep genetic causes are not just something GWASs can't detect, but in fact are not causes at all. Then, just two sentences later, M&H use the example of cystic fibrosis as a genetic cause that they say is uniform, unitary, and explanatory, that is, a deep genetic cause (target article, sect. 3.3, para. 5).
We propose that this inconsistency on the status of deep genetic causes arises from the account of causation M&H defend. As noted earlier, their account of causation is meant to cover both genetic causes and shallow genetic causes, which allows them to claim at one point that GWASs can't possibly be giving fuel to eugenicists because genetic causes by definition are non-uniform, non-unitary, and non-explanatory (target article, sect. 1.3, para. 1 & sect. 3.5, para. 1). So, what are we to make of deep genetic causes? M&H clearly want to claim that some genetic causes are shallow and some are deep, and that GWASs can only find shallow genetic causes. In fact, M&H's distinction between shallow genetic causes and deep genetic causes seems to be set up to explicitly acknowledge that there are various kinds of genetic causes. If this is right, then they need a different account of genetic cause, and possibly cause in general, in order to accommodate both shallow and deep genetic causes.
M&H briefly review a few philosophical accounts of causation, then claim to adopt a probabilistic version of Jim Woodward's (Reference Woodward2003) manipulationist account. However, it is unclear how this account fits GWASs, because there is no careful manipulation of specific variables in GWASs, a point M&H acknowledge when arguing for the superiority of within-family studies over regular GWASs (target article, sect. 3.5, para. 1). Furthermore, the distinction between shallow versus deep causes does not straightforwardly map on to much philosophical discussion of genetic causation (see e.g., Gannett, Reference Gannett1999; Lynch & Bourrat, Reference Lynch and Bourrat2017; Noble, Reference Noble2008; Northcott, Reference Northcott, Plaisance and Reydon2012; Oftedal, Reference Oftedal2005; Schaffner, Reference Schaffner2016; Sober, Reference Sober, Buchanan, Brock, Wikler and Daniels2000; Waters, Reference Waters2007). For example, M&H say thatcauses need not be mechanistic (target article, sect. 3.3, para. 11), while many philosophers, including those cited here, offer accounts ofgenetic causation as being necessarily mechanistic.
M&H's shallow versus deep distinction better tracks Eric Turkheimer's (Reference Turkheimer1998, Reference Turkheimer2016) distinction between strong and weak genetic explanations than it does distinctions in the philosophy of science literature on genetic causation. For Turkheimer a weak genetic explanation says “one way or another, genetic differences among people wind up correlated with phenotypic differences” (Turkheimer, Reference Turkheimer2016, p. 24). Neither shallow genetic causes nor weak genetic explanations determine the exact difference that certain single-nucleotide polymorphisms (SNPs) make in a phenotype but this is by design. A strong genetic explanation by contrast is “the discovery that an observed phenotypic difference is a manifestation of a specific latent genetic mechanism” (Turkheimer, Reference Turkheimer2016, p. 25). Both uncovering deep genetic causes and providing strong genetic explanations can reveal specific mechanisms.
While M&H's shallow versus deep distinction may resolve some apparent contradictions in their claims about GWASs and causation, we propose that their distinction better points to a way of separating alternate types of genetic explanations. This allows that GWASs do lead to genetic explanations of traits but only in the same sense in which non-genetic social science leads to environmental explanations of traits.
Acknowledgment
The authors thank Hannah Allen for her comments on the manuscript.
Competing interest
None.