It Got Eaten
- BuyWhat Darwin Got Wrong by Jerry Fodor and Massimo Piattelli-Palmarini
Profile, 262 pp, £20.00, February 2010, ISBN 978 1 84668 219 3
In 1959 the psychological doctrine known as ‘behaviourism’ was at the peak of its influence. Pioneered in the early 20th century by Edward Lee Thorndike, Clark Hull and J.B. Watson, behaviourism rejected explanations of action in terms of mysterious inner processes such as ‘thought’ and tried to explain behaviour purely in terms of the organism’s conditioning by experience. By the middle of the century, the behaviourist approach had been developed in a detailed and radical form by B.F. Skinner. Skinner explained learning in terms of reinforcement: organisms produce novel behaviours spontaneously, and those that are positively reinforced are more likely to occur in similar circumstances in the future. This view, developed in work on rats and pigeons, was extended to cover human language in Skinner’s 1957 book Verbal Behaviour. A young linguist, Noam Chomsky, published a review of Verbal Behaviour two years later. It was perhaps the most devastating book review ever written.
Chomsky argued that Skinner’s theoretical vocabulary could be applied to human linguistic behaviour only in an empty, post hoc way. He also thought that Skinner’s behaviourism had a simple architectural flaw: it held that external factors – especially experiences of reinforcement – were of ‘overwhelming importance’ in the explanation of behaviour. Hardly any role was given to what Chomsky referred to simply as ‘the internal structure of the organism’. It is unusual to do serious damage to a scientific research programme with a set of general arguments – not by citing experimental or mathematical results, but by looking at the basic ideas and revealing a crack in the foundations. Though the impact of the review itself is sometimes exaggerated, this is the effect Chomsky had on the behaviourist study of humans.
Jerry Fodor now hopes to do something similar to Darwinism in biology. Fodor has been making sceptical remarks about Darwinian ideas for decades. Three years ago he wrote a direct attack on Darwinian evolutionary theory in the LRB,[*] and he has now published What Darwin Got Wrong, along with Massimo Piattelli-Palmarini. Fodor and Piattelli-Palmarini believe that they can replicate Chomsky’s demolition job on Skinner because ‘Skinner’s account of learning and Darwin’s account of evolution are identical in all but name.’ As we shall see, ‘identical’ is quite a stretch, but there is a real analogy. First, both theories draw on an externalist or ‘outside-in’ pattern of explanation, in which the structure or behaviour of living things is seen as a consequence of their environments. Second, both rely on a process that can be described loosely as ‘trial and error’. New variations are produced in a spontaneous and unintelligent way, and a few successful variants are kept while others are discarded. This resemblance between the theories was recognised by Skinner himself, who saw Darwinism and behaviourism as dovetailing together. And Chomsky, the great critic of behaviourism, shows no enthusiasm for mainstream evolutionary theory, especially when applied to psychological traits.
Opposing attitudes towards externalist explanations point to a fissure running through the history of science and philosophy. This is part of what distinguishes the ‘empiricist’ from the ‘rationalist’ tradition, with the former (Locke, for example) explaining the contents of the mind in terms of what is given in experience, and the latter (Leibniz) insisting on the self-propelled power of thought. Which side of the divide one places oneself on is partly a matter of intellectual temperament; but Skinnerian behaviourism and Darwinism are still scientific theories, answerable to empirical evidence. The partial similarity in pattern does not mean they must stand or fall together. They are about different things; one may be right and the other wrong.
Modern evolutionary biology contains two main sets of theoretical ideas. One is the hypothesis of a pattern of common ancestry, roughly tree-shaped, linking all life on Earth. The other concerns the way change occurs within particular populations – or particular segments of life’s ‘tree’. Fodor and Piattelli-Palmarini have no argument with the claim of common ancestry. Their concern is with evolution by natural selection as a process of change within populations.
Suppose we have a population, i.e. a collection of organisms of some particular species. They live, reproduce and die. New variations appear from time to time, and do so in an ‘undirected’ manner. This variation can be loosely referred to as ‘random’, but it is due to ordinary physical causes. What is important is that the appearance of a new variant is not affected by whether or not it is likely to be useful. Most of these new traits are harmful, but a few will help organisms to survive and reproduce. If, in addition, these favourable traits are inherited, they may spread through the population.
This causal schema is essentially owed to Darwin, but since his time it has seen much elaboration and refinement, mostly thanks to the newer science of genetics. For some biologists, the idea of launching an attack on Darwin today is about as sensible as a physicist launching one on Galileo, so much has evolutionary biology changed since Darwin’s time. But Fodor and Piattelli-Palmarini are attacking ideas in evolutionary theory so basic that, for the most part, they go back to Darwin.
Even in the simple sketch I’ve just given, we can see ways in which the picture of Darwinian biology as an externalist or environment-driven theory is only partly right. Evolutionary success is a matter of survival and reproduction. A lot of what makes for an organism’s survival and capacity for reproduction depends on its relation to its environment, but a lot is a matter of the internal functioning of the organism’s machinery. And though organisms adapt to their environments, they also transform them: the oxygen-rich atmosphere of the Earth is a product of the evolution of life. In the first half of their book, Fodor and Piattelli-Palmarini give a diverse list of evolutionary factors that they see as being at odds with Darwinism. These all have to do with the ‘internal’ structure of organisms. But it is far from true that whenever anything internal proves to be important, this is non-Darwinian. Some of the more speculative ideas Fodor and Piattelli-Palmarini discuss create tensions with Darwinian views, while others (the ‘imprinting’ of genes, for example) are embraced by the most uncompromising Darwinians.
In the correspondence following Fodor’s article in the LRB, several critics pointed out that modern evolutionary biology makes use of a mix of explanatory factors, and puts particular emphasis on the evolutionary consequences of detailed features of genetic systems. (Darwin, for that matter, often gestured towards the importance of internal features of inheritance mechanisms, though he knew little about them.) In reply, Fodor has insisted that his targets are those evolutionists resolutely committed to giving explanations of living things in terms of ‘exogenous’ (external) factors. He sees any invocation of internal causes as a concession. But to say this is to set up an opponent very different from mainstream evolutionary biology, even evolutionary biology of the most vigorously and self-consciously ‘Darwinian’ kind.
The central aim of Fodor and Piattelli-Palmarini’s book is to undermine what is usually seen as a quite straightforward form of explanation in evolutionary biology, in which the distribution of features in a population at some point in time is explained in terms of the operation of natural selection. Such explanations take for granted that there is (or was) a particular range of variants in the population, and aim to explain why some are common, some are rare, some may have been lost altogether and so on.
A simple example: suppose we have a population of organisms and a new trait, T, appears in it. Individuals with T are more likely to survive to reproductive age, so on average they have more offspring. As T tends to be inherited, over time it increases in frequency in the population. It seems natural to say that there was selection for T. But what if there is another trait, T*, that organisms with T tend also to have? This might be because the genes that produce T also produce T*, or because the genes that produce T are close to those that produce T* in the genome and tend to be passed on together, or for other reasons. However the correlation arises, when explaining the increase in what we are calling ‘organisms with T’, we have to say why it is T that selection was favouring. Take the extreme case in which all organisms with T also have T*, and all with T* have T. Then it seems that natural selection is blind to the difference between them. So, Fodor and Piattelli-Palmarini say, whenever there is this kind of correlation between traits (it is ubiquitous, in various degrees), an explanation of the form ‘the population is now the way it is because trait T was favoured by natural selection’ is undermined.
They cast their argument within a framework derived from logic and the philosophy of language. There, two terms are said to be ‘co-extensive’ when they pick out exactly the same object or set of objects. Suppose that every animal that has a heart also has blood, and vice versa. Then the descriptions ‘animal with a heart’ and ‘animal with blood’ are co-extensive. Extending this usage, Fodor and Piattelli-Palmarini argue that two properties are co-extensive if all the same objects have those properties. So properties T and T* are co-extensive if everything with T also has T*, and everything with T* also has T. Two properties can also be ‘locally co-extensive’ if in some region everything which has one property also has the other, even if the properties are not co-extensive in all places and times. In a more standard way of talking, we might say that the properties are ‘correlated’ (either locally or everywhere).
‘Intensionality’ is another term of art required here. A sentence is intensional if it can be changed from true to false, or from false to true, by substituting one co-extensive term in it for another. It might be true that ‘Bob thinks the morning star is bright’ but false that ‘Bob thinks the evening star is bright’ even though ‘morning star’ and ‘evening star’ refer to the same object. Sentences containing ‘thinks that …’ are often intensional. Fodor and Piattelli-Palmarini claim that an ‘intensional process’ is one that ‘distinguishes’ between co-extensive properties, and that Darwinism requires natural selection to be an intensional process, where in fact it is not. Darwinism is thus guilty of an ‘intensional fallacy’.
It is easier to see what is going on here by putting it in simpler terms. Continuing with the example above: some individuals in our population have T, and these also have T*. As selection is a contrastive matter, we also need to label an alternative form: let’s say other individuals have U rather than T, and that these have their own correlated trait, U*. Let’s now suppose that T and U are alternative colours of the exterior of the organism, and T* and U* alternative colours of some interior part of the organism that is never normally seen. The colour of this interior structure is determined as a by-product of the processes determining the exterior colour (perhaps it is the same colour, perhaps a different one).
Next, it is helpful to switch from thinking about selection for a trait to its flipside, selection against. If T and U are alternative colours of the outside of the organism, suppose that those with U are more often eaten by predators. If we look at some particular individual with U, does it make sense to ask whether it was U, U* or both that led to its dying young? We ask this setting aside questions about evolutionary theory – our interest is just in this one organism and its untimely death. Does it make sense to say that the death was due to one of the correlated properties and not the other? Surely it does. We work this out by looking at the processes in which U and U* are involved – the biochemical pathways, the roles that different parts of the animal play in its life. This individual was seen by a predator and eaten, and it was seen because the colour of its exterior meant it was poorly camouflaged. The fact that an internal part of the organism had a particular colour was irrelevant.
If this can be true of one individual, it can be true of ten of them, or a thousand. So we have a population in which individuals with U and U* are dying young while individuals with T and T* are not. For each of those with U and U*, it is U that is hastening their demise while U* is irrelevant. Similarly, for each of those with T and T*, T helps while T* is irrelevant. Add to which that T has no other costs that are not shared with U. This then is a case of natural selection. If the combination of T and T* is inherited over generations, then (barring special factors interfering) it will increase in frequency in the population. T was selected for, and U selected against, because these are the features that can determine whether individuals live or die. T* increases in frequency along with T, but it is just along for the ride.
Perhaps there is something strange in this way of putting things, whereby a trait like T* can be made more common by natural selection, without being selected for. One may wonder whether an implication of agency isn’t lurking in the term ‘selection’. That is in fact what Fodor and Piattelli-Palmarini claim. In which case, the thing to do is to set the suspicious terms aside and see whether it is possible to describe the same sequence of events without using them. If it is, the theory of natural selection is not called into question, and at most there is an awkwardness in the terminology. This turns out to be the case. Suppose we say, for example, that individuals with T and T* had more offspring than individuals with U and U*, and did so because T helped them avoid predators. T* also increased in frequency, but only because it is passed on along with T. However we choose to talk about it, the theory does distinguish between two different cases, one where a trait affects survival and reproduction and another where it doesn’t. In practice, confusion arising from the unwanted connotations of the term ‘selection’ is rare in biology, but it is always possible to shift to a more careful description if they intrude.
In the light of this, let’s look more closely at what Fodor and Piattelli-Palmarini say about causal relations. Here some further philosophical machinery is used. What might we say to flesh out the claim that selection was favouring T in my example above? Perhaps something like this: if the organisms with T did not have T*, they would still have been more successful than those with U, and if they had T* but not T, they would not have been more successful. An ‘if … then’ claim which is at odds with how things actually are constitutes a ‘counterfactual’. Fodor and Piattelli-Palmarini argue that ‘the distinction between traits that are selected for and their free-riders turns on the truth (or falsity) of relevant counterfactuals,’ but that a selection process is not affected by how things merely could have been. ‘Selection cannot, as a matter of principle, be contingent upon (merely) counterfactual outcomes.’
Counterfactuals are a popular tool in philosophy when discussing causal relations. They can be helpful, but they can also be part of a rather misleading picture: sometimes people write as if one should try to inspect or survey the counterfactuals that apply in a particular case, and treat them as a kind of litmus test for causation. Fodor and Piattelli-Palmarini suppose that you have to be able to do this before you can determine whether T or T* is what was selected for. But for T rather than T* to be ‘selected for’ just means that T is the trait that gave individuals their advantage. We can know this by looking at what T does, and how it relates to the organisms’ lives. T* and U*, on the other hand, have no role in the organisms’ lives and deaths. Having discovered this, we may then feel confident in asserting a counterfactual: if the successful organisms had T but not T*, they would still have done well. This is in fact a post hoc summary of the causal knowledge we gained by looking at the biology of the situation – at what the traits in question actually do.
In the discussion following both Fodor’s 2007 article and the publication earlier this year of What Darwin Got Wrong, a number of commentators have argued that there is no conceptual problem for evolutionary theory arising from correlated traits, as one of the traits can be important in causing reproductive success while the other is not. Fodor and Piattelli-Palmarini have suggested in reply that their argument has been misunderstood. In response to Ned Block and Philip Kitcher, for example, they say it was never their intention to claim that there cannot be a fact of the matter about which of two correlated traits ‘causes increased reproductive success’. Rather, their claim is that even if one trait causes organisms to reproduce more while the other trait is irrelevant, this is not a situation in which natural selection is distinguishing between the two. For that, something more would be needed, perhaps a special kind of law of nature which they call a law of ‘adaptation as such’. But, again, if one trait is causing increased reproductive success, it is being selected for, in the sense that matters to evolutionary theory. Nothing more is needed. Fodor and Piattelli-Palmarini criticise the tendency to talk of selection as if it were an agent. They are right that this is often misleading, but they seem to be making a similar mistake when they treat it as something over and above the ordinary facts of life, death and reproduction. For Fodor and Piattelli-Palmarini, it makes sense to ask: ‘Even if trait T causes organisms to reproduce more while T* has no effect, how can selection see that fact?’ But there is no question to ask here, nothing extra that selection might achieve or fail to do.
Given that Fodor and Piattelli-Palmarini do not show any real flaw in the architecture of Darwinian biology, let me return to the relation between behaviourism and Darwinism. The similarities between behaviourist learning theory and Darwinian evolutionary theory are not as close as Skinner or, now, Fodor and Piattelli-Palmarini claim, even if the theories have structural resemblances. Behaviourist learning theory, as applied to humans, has been largely abandoned, but that doesn’t mean that Darwinism’s theory asks to be abandoned along with it. Still, if one view is right and the other wrong, we might try to say something about why a certain pattern of explanation works in one place and not in the other.
Darwinian processes of variation and selection have built our brains. These processes have constructed a mechanism that goes through a kind of adaptive change – we call it ‘learning’ – within each person’s lifetime. But evolution was not bound to build brains which use the same mechanism of change that we see in evolution itself. One Darwinian process can sometimes build another Darwinian machine running on a faster time-scale: one part of our immune system, for example, runs on a variation-and-selection process in which many novel antibody molecules are produced in an undirected way, and the cells that happen to make antibodies which can recognise invaders become more numerous. But this is not, for the most part, what evolution has done in the building of our brains. There is some role for simple ‘trial and error’ in learning, but much of our learning is smarter than that. We can produce new behaviours because they seem likely to work. We can learn by watching. We, unlike our genes, can reason through problems. Our behaviours reflect much more than experiences of ‘reinforcement’, and that is why Skinner was wrong while Darwin was right.