In arguments about innateness, one often finds a bias towards empiricist perspectives. It is widely, though erroneously, believed that if some aspect of cognition could be learned, that aspect of cognition must be the product of learning; evidence for the possibility of learning is often taken as evidence against possibility of innateness. Of course, in reality, some aspects of cognition could be innate, even if they were in principle learnable. Humans might, in principle, be able to learn how to walk, much as they can acquire other new motor skills (e.g., juggling or skiing), but the fundamental alternating stepping reflex that underlies walking appears to exist at birth, prior to any experience of actual walking.
In a similar way, in discussions about adaptive advantage, it sometimes seems as if there is a bias towards adaptationist accounts relative to by-product accounts, such that any putative adaptive advantage apparently automatically trumps the possibility of non-adaptive accounts. If something could have been shaped by adaptive pressure, it is often assumed to have done so; but it is again a logical error to assume that simply because something could be explained as an adaptation then it could not also be explained in another fashion.
McKay & Dennett (M&D) make a reasonably strong case for the possibility that positive illusions, such as non-veridical beliefs about one's health, could have a history borne of direct adaptive advantage; is there any reason to consider alternative accounts? Quite possibly: Although positive illusions might inhere in some sort of domain-specific cognitive substrate that could have been specifically shaped via natural selection, I believe it is at least equally plausible that positive illusions are simply one manifestation among many of two considerably more general phenomena that are pervasive throughout human cognition (yet curiously absent from the target article): (1) confirmation bias (e.g., Nickerson Reference Nickerson1998), in which people often tend to cling to prior beliefs even in the light of contradictory evidence; and (2) motivated reasoning, a tendency of people to subject beliefs that are potentially ego-dystonic to greater levels of scrutiny that are less likely to be ego-dystonic (Kunda Reference Kunda1990).
Cigarette-smokers, for example, tend to dismiss research on the dangers of smoking not because they believe themselves to be healthy, but because they work harder to deflate potentially damaging beliefs. This could be seen as an instance of a cognitive mechanism that was dedicated towards yields positive self-illusion. But much the same tendency towards motivated reasoning can be seen in undergraduates' evaluations of arguments about capital punishment: As Lord et al. (Reference Lord, Ross and Lepper1979) showed, students tend to work harder to undermine arguments that conflict with their prior beliefs. Dozens of subsequent studies, reviewed in Kunda (Reference Kunda1990), point in the same direction: We work harder to dismiss arguments that we don't like, whether or not those arguments pertain to our own personal well-being. Positive illusion might in this way be seen as an instantiation of motivated reasoning, rather than as the product of a dedicated mechanism with a unique adaptive history.
A species-general tendency towards confirmation bias might also subserve positive illusions, even absent specific machinery dedicated to positive illusion. Once one stumbles on a belief in one's own virtues (e.g., through the praise of one's parent), potentially disconfirming evidence may be ignored, underweighted, or simply harder to retrieve; but again, there may be nothing special about personal self-interest. Confirmation bias is as apparent in people's inferences about arbitrary rules in concept learning tasks as it is in beliefs about self (again, see Nickerson Reference Nickerson1998, for a review).
Intriguingly, confirmation bias itself may be a byproduct of the organization of human memory (Marcus Reference Marcus2008; Reference Marcus2009). Human memory, like that of all vertebrate creatures, is organized via context rather than location. In a machine with location-addressable memory, it as easy to search for matches to be a particular criterion (data that fits some theory) as data that does not match said criteria (potentially disconfirming evidence). By contrast, in a creature that can search only by contextual matches, there is no clean way in which to access disconfirming evidence. Confirmation bias itself may thus stem from inherited properties of our memory mechanisms, rather than specific adaptive advantage. Indeed, as M&D themselves argue, an across-the-board bias towards misbelief per se is unlikely to be adaptive.
At the present time, we simply lack the tools to directly infer evolutionary history. Some adaptationist theories are likely to be correct – parental investment theory, for example, is supported by a vast range of supporting evidence; others, such as the theory that depression serves to keep its bearers from getting into trouble, seem rather more dubious. Advances in understanding how cognitive machinery is instantiated in underlying brain matter may help, as may advances in relating genetic material to neural structure; for now, we are mostly just guessing. Are positive illusions in fact evolved misbeliefs or are they merely by-products of more general mechanisms? We really can't say. My point is simply that we should be reluctant to take either option at face value. Specific properties of cognitive machinery may sometimes turn out to be by-products even when it superficially appears in principle that there is an adaptive pressure that could explain them.
