2500 words

It’s been almost 5 years since I read What Darwin Got Wrong (WDGW) (Fodor and Piattelli-Palmarini, 2009; F&PP) which changed my view on the theory of natural selection (ToNS). In the book, they argue that natural selection cannot possibly be a mechanism since it cannot distinguish between correlated traits, since there is no mind (agent) doing the selecting nor are there laws of selection for trait fixation across all ecologies. Fodor had originally published Why Pigs Don’t Have Wings in the London Review of Books in 2007, and then he published Against Darwinism (Fodor, 2008) where he mounted and formulated his argument against the ToNS.

A precursor to the argument against the ToNS

Although Fodor had begun articulating his argument in the late 2000s, he already had a precursor to it in a 1990 paper A Theory of Content (Fodor, 1990: 72-73):

The Moral, to repeat, is that (within certain broad limits , presently to be defined) Darwin doesn’t care how you describe the intentional objects of frog snaps. All that matters for selection is how many flies the frog manages to ingest in consequence of its snapping, and this number comes out exactly the same whether one describes the function of the snap-guidance mechanisms with respect to a world that is populated by flies that are, de facto, ambient black dots, or with respect to a world that is populated by ambient black dots that are, de facto, flies.19 “Erst kommt das Fressen, denn kommt die Morale.” Darwin cares how many flies you eat, but not what description you eat them under.(Similarly, by the way, flies may be assumed to be indifferent to the descriptions under which frogs eat them.) So it’s no use looking to Darwin to get you out of the disjunction problem.

In Against Darwinism and WDGW, F&PP reformulate and add to this argument, stating that it is a selection-for problem:

In a nutshell: if the assumption of local coextensivity holds (as, of course, it perfectly well might), then fixing the cause of the frog’s snaps doesn’t fix the content of its intention in snapping: either an intention to snap at a fly or an intention to snap at an ABN would be compatible with a causal account of what the frog has in mind when it snaps. So causal accounts of content encounter a selection-for problem: If something is a fly if and only if it is an ABN, the frog’s behaviour is correctly described either as caused by flies or as caused by ABNs. So, it seems, a causal theory of content cannot distinguish snaps that manifest intentions to catch the one from snaps that manifest intentions to catch the other. (Fodor and Piattelli-Palmarini, 2010: 108)

The argument

Fodor has formulated the argument two times, in Against Darwinism (Fodor, 2008: 11-12) and then again in WDGW (Fodor and Piattelli-Palmarini, 2010: 114).

Contrary to Darwinism, the theory of natural selection can’t explain the distribution of phenotypic traits in biological populations.

(i) To do so would require a notion of ‘selection for’ a trait. ‘Selects for….’ (unlike ‘selects…) is opaque to substitution of co-referring expressions at the ‘…’ position.

(ii) If T1 and T2 are coextensive traits, the distinction between selection for T and selection for T2 depends on counterfactuals about which of them. The truth makers for such counterfactuals must be either (a) the intensions of the agent that affects the selection, or (b) laws that determine the how the relative fitness of having the traits would be selected in a possible world where the coextension does not hold.

(iii) But:

Not (a) because there is no agent of natural selection.

Not (b) because considerations of contextual sensitivity make it unlikely that there are laws of relative fitness (‘laws of selection).

QED (Fodor, 2008)

---

1. Selection-for is a causal process.
2. Actual causal relations aren’t sensitive to counterfactual states of affairs: if it wasn’t the case that A, then the fact that it’s being A would have caused its being B doesn’t explain its being the case that B.
3. But the distinction between traits that are selected-for and their free-riders turns on the truth (or falsity) of relevant counterfactuals.
4. So if T and T’ are coextensive, selection cannot distinguish the case in which T free-rides on T’ from the case that T’ free-rides on T.
5. So the claim that selection is the mechanism of evolution cannot be true. (Fodor and Piattelli-Palmarini, 2010: 114)

Selection for

Ernst Mayr wrote in One Long Argument that “Selection-for specifies the particular phenotypic attribute and corresponding component of the genotype (DNA) that is responsible for the success of the selected individual.” Selection-for a trait is needed for adaptationism to be true. But what kinds of hallmarks of adaptation are there, that a free-riding trait that’s not an adaptation would have? Selection-for problems need to appeal to counterfactuals, and to appeal to counterfactuals, they need laws of relative fitness. Selection-for problems arise when so-called explanations require distinguishing the causal role of coextensive properties (traits) (Fodor and Piattelli-Palmarini, 2010: 111).

But for there to be counterfactual-supporting laws that would be able to distinguish the correlation and select the fit trait over the free-riding trait, there need to be laws that apply across all ecologies and phenotypes. But the case of whether or not T1 or T2 is conducive to fitness is massively context-sensitive. So there can’t be laws of relative fitness. T may be helpful in one ecology and not another. T may also be helpful for survival in one organism in a specific ecology but not another organism. Therefore, there is no law that explains why T would win a trait competition over T2; so there aren’t any laws of relative fitness. The ToNS implies a filtering mechanism which is the environment. But the environment can only access the correlation, and it explains the selection of both traits without being able to say anything about what’s selected-for.

A mechanism?

What is a causal mechanism? A causal mechanism is a sequence of events or processes governed by lawlike regularities.  Basically, causal claims are claims about the existence of a mechanism. In the ToNS instance, the lawlike regularities would refer to laws of relative fitness. In the Fodorian sense, the mechanism would need to be sensitive to correlated traits. But the only thing that would be sensitive would be laws of relative fitness. So the question is, how can natural selection be a mechanism if it can’t ground counterfactuals which distinguish selection of from selection-for? Laws would support counterfactuals. The kind of law that Fodor is looking for is “All else being equal, the probability that a t1 wins a competition with a t2 in ecological situation E is p.” Basically, if there are laws they can support the counterfactuals. However, due to massive context-sensitivity, it seems unlikely that there are ceteris paribus laws of relative fitness. If there are, why hasn’t anyone articulated them?

But the fact of the matter is this: Fodor has successfully argued against the claim that natural selection is a mechanism, and he is not even alone on that, since others argue that natural selection isn’t a mechanism, not using his arguments (eg Skipper and Millstein, 2005; Havstad, 2011).

The question is: How can natural selection be a mechanism if it can’t ground the counterfactuals that distinguish selection of from selection for? Because if T and T’ are correlated, the same story explains the selection of both traits, so we can’t use the ToNS to show which was selected-for its causal contributions to fitness and which was the free-riding trait that just came along for the ride. Thus, the ToNS doesn’t explain the trait and so if it doesn’t explain the trait then it doesn’t predict the trait. Natural selection makes no prediction as to which trait will be selected-for. That’s not to say that we (humans) can’t know what was selected-for from what was merely selected, and Fodor never claimed otherwise, contrary to those who claim that he did make that claim. The best example is Pigliucci (2010) who states in his review that:

functional analyses rooted in physiology, genetics and developmental biology, and why observations of selection in the field are whenever possible coupled with manipulative experiments that make it possible to distinguish between [correlated traits].

Fodor was emphatic about this—he never claimed that humans couldn’t distinguish between correlated traits, only that the, using the ToNS, we can’t know which trait was selected-for its contribution to fitness since, due to the correlation, the same story explains both traits. Fodor and Piattelli-Palmarini (2010) in Replies to Our Critics stated as much:

Many of the objections that have been raised against us seem unable to discriminate this claim from such quite different ones that we didn’t and don’t endorse, such as: when traits are coextensive, there is no fact of the matter about which is a cause of fitness; or, when traits are coextensive, there is no way to tell which of them is a cause of fitness; or when traits are coextensive Science cannot determine which is a cause of fitness…etc. Such views are, we think, preposterous on the face of them; we wouldn’t be caught dead holding them. To the contrary, it is precisely because there is a fact of the matter about which phenotypic traits cause fitness, and because there is no principled reason why such facts should be inaccessible to empirical inquiry, that the failure of TNS to explain what distinguishes causally active traits from mere correlates of causally of active traits, shows that something is seriously wrong with TNS.

In his 2018 book Agents and Goals in Evolution, Samir Okasha (2018) states that Darwin was the first to employ what he terms “type II agential thinking”, which is to personify “mother nature” as selecting fit traits. Darwin’s analogy between natural and artificial selection fails, though. It fails because in the case of artificial selection, there is a mind attempting to select fit traits while in the case of natural selection, there is no mind, so the only way around this is laws of relative fitness.

Gould and Lewontin (1979), although criticizing adaptive explanations, also held that natural selection is the most powerful mechanism of evolution. This claim—that natural selection is a mechanism—is ubiquitous in the literature. And it is Gould and Lewontin’s Spandrel argument that partly inspired the correlated/coextensive trait argument devised by Fodor. However, as F&PP note, Gould and Lewontin didn’t take their argument to its logical conclusion, which of course was the rejection of Darwinian natural selection in explaining the fixation of traits in organisms.

What are selected are not traits, but organisms. And just because an organism with T was selected, doesn’t mean that T was the cause of fitness. We can then say that the phrase “survival of the fittest” is a tautology, since the fit are those who survive and those who survive are fit. But Hunt (2014) claims that we can reformulate it: that it should be defined as a theory that attempts to predict and retrodict evolutionary change which acts upon organisms through the environment. However, this reformulation runs right into Fodor’s argument, since there is no way for the exogenous selector (the environment) to distinguish between correlated traits. Portin (2012) claims that we can reduce the tautology to “those who reproduce more, reproduce more”, stating that it merely “seems” like a tautology. Even Coyne and Dawkins, in their books Why Evolution is True and The Greatest Show on Earth make the mistake multiple times in explaining natural selection in a tautologous way (Hunt, 2012). The fact of the matter is, natural selection is nothing more than an oxymoronic tautology (Brey, 2002).

To explain something means to identify a causal mechanism for that explanation. In the case of natural selection, if we are to explain why T1 over T2 was selected-for, we would then need to identify the causal mechanism that would be able to distinguish between correlated traits. The ToNS is a probabilistic theory. If NS is to be a mechanism, then since it is probabilistic, it has to be a stochastic mechanism. Though leading accounts of mechanisms are deterministic. Therefore, NS can’t be a mechanism (Skipper and Millstein, 2005). There are, as a matter of fact, no stable organizations of the component parts in NS, so it can’t be a mechanism.

Conclusion

Nanay (2022: 175) tells us that Fodor’s book was disregarded by many academic publishers, which then speaks to what Leal (2022) notes as the emotionality from Fodor’s detractors:

I knew about [What Darwin Got Wrong] long before the 2009 publication date (as early as 2005, as I recall), because all my friends and colleagues working in philosophy of biology kept telling me how they had to reject, in no uncertain terms, this book from all major and then not so major academic publishers. Fodor then took it to a non-academic publisher, and it did get published,

So the question is, how does the ToNS explain the trait if it doesn’t predict which of the two correlated traits move to fixation if it can’t discern between the two correlated traits? How does the ToNS predict which trait is fitness-enhancing when two traits are correlated prior to performing an experiment? That’s, again, where humans come in. We can perform experimental manipulations and then discern the fit trait from the correlated trait that does not cause fitness. But we can’t merely use the ToNS to do this.

There is also the fact that most if not all respondents to Fodor did not understand the argument, and this can be seen from their responses to him, like Pigliucci’s where he talks about experimentation. Humans can have access to the the fit trait, but the environment—the exogenous filter—only has access to the correlation, and so the same story explains both traits.

So neo-Darwinism is false and natural selection isn’t a mechanism. What, then, is the alternative? Fodor ended Why Pigs Don’t Have Wings writing:

The alternative possibility to Darwin’s is that the direction of phenotypic change is very largely determined by endogenous variables. The current literature suggests that alterations in the timing of genetically controlled developmental processes is often the endogenous variable of choice; hence the ‘devo’ in ‘evo-devo’.

“Endogenous variables” mean causes within the organism. For instance West-Eberhard (2003: 179) noted that: “If genes are usually followers rather than leaders in evolution—that is, most gene-frequency change follows, rather than initiates, the evolution of adaptive traits—then the most important role of mutation in evolution may be to contribute not so much to the origin of phenotypic novelties as to the store of genetic variation available for long-term gradual genetic change under selection.” So that is one way that endogenous variables can direct phenotypic change. This has been noted by many recent authors (eg Noble et al, 2014).

One of the Darwinian premises is that mutations occur randomly. However, we now know that this is not always the case: “genetic change is far from random and often not gradual” (Noble, 2013). All of the assumptions of neo-Darwinism have been disproved, most importantly, the theory of natural selection being the causal mechanism of evolution.

P1: Natural selection is a mechanism iff it can distinguish between causes and correlates of causes.

P2: Natural selection can’t distinguish between causes and correlates of causes.

C: Therefore, natural selection isn’t a mechanism.