What Fodor and Piattelli-Palmarini got wrong – part 2: the 'selection-for' problem

In my previous post I discussed some developments in evolutionary theory that Fodor and Piatelli-Palmarini thought were problematic for Darwinism. I saved their main argument for another post. So what is their main argument? In their own words:

Because selection-for is intensional (or, if you prefer, because what are selected-for are not creatures, but their traits, and the individuation of traits is intensional) there can be coextensive but distinct phenotypic properties, one (but not the other) of which is conducive to fitness, but which natural selection cannot distinguish. In such cases, natural selection cannot, as it were, tell the arches from the spandrels. That being so, adaptationist theories of evolution are unable, as a matter of principle, to do what they purport to do: explain the distribution of phenotypic traits in a population as a function of its history of selection for fitness.

Their point is that some traits are linked so that a change in trait t1 will go together with a change in trait t2. In these cases t1 might have been selected for and t2 might be a free-rider (i.e. does not influence fitness), or t2 might have been selected for and t1 might be a free-rider. Natural selection cannot distinguish these two alternatives. The basis for this argument comes from a paper by Gould and Lewontin (1979). In the above quote Fodor and Piatelli-Palmarini allude to this paper when they say that "Natural selection cannot, as it were, tell the arches from the spandrels." What do arches and spandrels have to do with natural selection? Gould and Lewontin make an analogy between the free-riding of spandrels when building arches for a dome. The spandrels are the triangular shaped spaces between adjacent arches.

The point is that when the arches are designed the spandrels will appear as well. You cannot have arches without spandrels. Likewise, in evolution two traits might be linked such that selection for trait t1 will also produce the free-rider t2. The lesson is that we need to be careful when we try to discover what a trait was selected for, because the trait might just as easily have been a free-rider. At no point do Gould and Lewontin deny natural selection, but Fodor and Piatelli-Palmarini actually argue that the argument undermines natural selection.

The authors seem to be somewhat confused about how natural selection is supposed to work:

An important consequence of genetic pleiotropism is that, when a gene affects several traits at once, any change in that gene that is not catastrophic (any viable mutation) will affect all or most of its traits. Supposing that one such change in one such trait is adaptive, then natural selection will eventually fixate that mutation. But then all the other changes in all the other traits will also be stabilized, possibly opening up wholly different selective processes, eventually dwarfing the effects of the initial selection driven by the initially adaptive trait.

This line of argumentation is quite baffling and seems to be a consequence of thinking of an intensional force called natural selection selecting for traits on the basis of fitness. If two traits are genetically linked then natural selection cannot select for one of them only to be affected by the other trait at a later time. The correct way of thinking about natural selection was explained by Richard Dawkins in The selfish gene: Selection occurs at the level of the gene. When a genetic mutation increases reproductive success this mutation will spread through the population. It is irrelevant how many traits are affected by the mutation: natural selection will work on the overall outcome and not the individual traits. Traits are not selected, but genes are.

Not surprisingly, the authors return to a comparison between natural selection and operant conditioning (their other archnemesis) to explain the problem. What if we train a pigeon to differentiate a yellow triangle from an X. Whenever the pigeon pecks at the yellow triangle it gets a food reward, but when it pecks at the X it does not. The pigeon will then learn to peck at the yellow triangle. But, ask Fodor and Piatelli-Palmarini, what has the pigeon really learned? Did it learn to peck at anything yellow? Did it learn to peck at a triangle? Did it learn to peck at a yellow triangle? Take a look at this video if you will to see operant conditioning of a pigeon in action:

 



What has the pigeon learned? Has it learned to peck at the word "PECK" and to turn at the word "TURN"? Has it learned to peck at letter "P" or "E" or "C" or "K" and to turn at the letter "T" or "U" or "R" or "N"? Fodor and Piatelli-Palmarini think this is a problem and that the only way to solve this problem is to examine counterfactuals. For example, we could present the letter "T" without the whole word "TURN" and see what happens, or we could present "URN" and see what happens. Why is this relevant for natural selection? Well, just like the letter "T" and the word "TURN" are coextensive in the operant conditioning experiment, some traits are coextensive in evolution by natural selection. The problem, according to the authors is that we cannot examine counterfactuals, because we cannot examine t1 without t2 if t1 and t2 are coextensive. This is perfectly true, but I think the analogy shows that they are missing the essence of both operant conditioning and natural selection. While it is true that we cannot distinguish between learning "T" and learning "TURN" on the basis of this single experiment, it should be noted that this is not what researchers are interested in when they do these experiments (at least not primarily). There are many interesting and important things we can learn about operant conditioning without caring at all about the actual stimulus. For example, as Skinner explains in this video, we can learn about how different reinforcement schedules affect behavior. Whether they learn the letter "T" or the word "TURN" the effect of reinforcement schedules will be similar and this is what matters. This effect generalizes to other species including humans and we can learn about gambling addictions on the basis of these experiments.

Likewise, the phenomenon of linkage is really not the most important aspect of natural selection. Yes, some traits are genetically linked, but this in no way challenges natural selection. Darwin argued that there is (1) descent with modification and (2) evolution by natural selection. Indeed, we know that mutations occur and this underlies descent with modification. Regardless of what the modification is and regardless of how many traits are affected by the mutation this modification is then subjected to natural selection. We know that phenotypic differences can result in different reproductive success. This is natural selection and it is a fact. It may not be the only way evolution occurs (we also have genetic drift for example), but it certainly plays an important role in evolution.

I will conclude this post with a quote from Fodor and Piatelli-Palmarini concerning what I think is their real interest in their attack against natural selection:

...since the Otxi 'master' gene controls the development of the larynx, inner ear, kidneys and external genitalia and the thickness of the cerebral cortex, selective pressures sensitive to changes in the function of the kidneys (due to bipedal station, or different liquid intake and excretion resulting from floods or droughts), or the fixation of different sexual patterns, may have in turn secondary effects on the expansion of the cerebral cortex and the structure of the larynx. The peculiarity of the overall picture of the evolution of language and cognition in humans, should this reconstruction prove to be correct, has been stressed to us by Boncinelli (personal communication).

This speculative hypothesis by Boncinelli was precisely what Fodor and Piatelli-Palmarini were looking for, because of their disapproval of the field of evolutionary psychology. They do not like the idea that cognitive mechanisms in humans were adaptations, so any idea that the increased brain size in humans was 'free-riding' on a different adaptation is music to their ears. There are two points to stress here: First, even if this were true it is still completely consistent with natural selection. 'Selective pressures sensitive to changes in the function of the kidneys...' is natural selection in action. So perhaps the authors aren't really challenging natural selection per se, but the view that every single trait must be an adaptation. Well, linkage is a well-known concept in evolutionary biology, so if this is their point they are merely beating a straw man. Second, even if the linking of these features were really as simple as is proposed here (surely there are other genes involved in addition to otxi) then natural selection operates on the combined effect of all these changes. Since there are reasons to believe that an increased brain size (a development lasting a few million years mind you) gave our ancestors survival benefits, it is unlikely that this would have been a 'free-riding' effect. Rather, it would have contributed, together with the other changes (e.g. in the kidneys), to the spread of the underlying genetic mutation.