2. Manufacture and malfunction

First consider the domain of systems designed and manufactured by humans. Here we envisage the distinction as one (codified in warranty legislation) between “culpable design limitations” (malfunctions) and “forgivable design limitations/features.” When a given artifact fails to perform a particular task, this failure is always due to a limitation in the design of that artifact. The question is whether the design limitation concerned is – from the designer's perspective – a blameworthy, “culpable” limitation (a design flaw, or perhaps a flaw in the execution of the design), or whether it is a tolerable, “forgivable” limitation. Examples of the former (with respect to the arbitrary task of “keeping time”) include:

1. 1. My $20,000 Bolex watch loses 10 seconds every day (contra the advertisement).

2. 2. My cheap Schmasio watch loses 10 minutes every day (contra the advertisement).

Examples of the latter limitation include:

1. 1. My toaster does not keep time at all.

2. 2. My Bolex loses a second every day (within warranted limits).

3. 3. My cheap Schmasio loses a minute every day (within warranted limits).

4. 4. After putting it in a very hot oven for an hour, my Bolex does not keep time at all.

What we can see from these examples is that manufactured artifacts either work as intended (within a tolerable margin of error), or they don't work as intended (falling outside the tolerable margin). What's important is not how well the artifacts objectively work, but how well they work relative to how they were intended to work (and the intentions of the manufacturer will bear upon the advertised claims of the manufacturer). The Bolex and Schmasio examples reflect this, because the malfunctioning Bolex still works objectively better than the properly functioning Schmasio.

Of course, some apparent design limitations are in fact deliberate design features. To cite a single example, contemporary consumers are frequently frustrated by DVD region code restrictions. The fact that a region 1 DVD player (sold in North America) cannot play discs sold in Europe or Japan (region 2) is certainly, from the consumer's perspective at least,Footnote ⁴ a limitation in the design of that DVD player – and often a frustrating limitation. In our terminology, however, the limitation is forgivable because such players are not designed to play DVDs from other regions, and indeed are deliberately designed not to do so. Region restrictions are, as software designers often say, “not a bug but a feature” of such machines, ostensibly to safeguard copyright and film distribution rights.

The essential lesson is that a manufactured artifact functions properly if it functions as its designer intended (and warranted) it to function, under the conditions in which it was intended (and warranted) to function. If the artifact fails to function under those conditions, then it has malfunctioned, which may be due to a flaw in the design or a flaw in the execution of the design. Here “malfunction” is equated with “culpable design limitation” and is defined so as to exclude seeming breaks in function that occur outside the constraints specified by the manufacturer (i.e., if a watch falls to pieces a day after the warranty expires, this is not a malfunction – not on our definition of malfunction, anyway – but a forgivable limitation).

Consider another example: Imagine a computer that is equipped with software for solving physics problems. The computer takes the problems as input, and produces purported solutions to the problems as output. Suppose, further, that the program that the computer implements when solving the problems utilizes Newtonian physics. Consider then three different possible scenarios:

1. The computer is assigned a problem about an apple falling from a tree on Earth. It produces the correct solution.

2. The computer is assigned a problem about an apple falling from a tree on Earth. Unfortunately, a low-level glitch occurs – a flaw in the execution of the program's design – causing the program to malfunction and to produce an incorrect solution.

3. The computer is assigned a problem about the mass of an apple as it approaches the speed of light. The program runs smoothly and predictably but arrives at an incorrect solution.

Do the second and third scenarios here map onto the distinction between culpable and forgivable design limitations? Whether this is the case depends on the precise intentions of the program designer. If the designer had implemented a Newtonian program because it was easier and cheaper to do so, but was fully aware that Einsteinian problems would compute incorrectly, then the third limitation is forgivable (if it was so advertised). If, however, the designer intended his or her program to solve physics problems of all types, then this limitation is culpable (and constitutes a malfunction, in this rather peculiar sense of the word).

Even such a common artifact as an electronic hand calculator produces output that may appear culpable:

A manufactured object (or feature thereof) works as a model of adaptive misbelief if: (1) The object is a specific focus of deliberate design (not a mistake or a by-product); (2) the object appears, from a certain perspective, to be malfunctioning or limited insofar as it misrepresents information to the consumer of that information; and (3) such misrepresentation is actually beneficial to the consumer of that information. None of the cases of artifacts considered thus far would qualify as analogues of adaptive misbelief under these criteria, but here is one case that gets close: the automotive mirror that is designed such that objects appear farther away than they really are. That this is misrepresentation is made clear by the appended cautionary subtitle (required, no doubt, by the manufacturer's lawyers): “objects in mirror are closer than they appear.” The trade-off in the goal of this design is clear: to provide a wider field of view than a “veridical” mirror, which is deemed a benefit that outweighs the distortion, a cost that is diminished, presumably, by the attached warning. The reason this example ultimately fails as a model of adaptive misbelief is that the misrepresentation itself is not the specific focus of design, nor is it (in and of itself) beneficial to the consumer; rather, the misrepresentation is an unavoidable by-product of producing a wider field of view.

We don't know of other good candidates but can describe a possible device with a similar design rationale: an alarm clock designed to set itself 10 minutes ahead in the middle of the night (and then to repair its “error” later in the day). Its design rationale would be to give its owner a little extra time to get going, but once the user figured this out, the device would of course lose effectiveness – a case of “the boy who cried wolf,” a design complication that we will discuss in some detail below. Before we move on, we note that whereas artifacts designed to misrepresent information to their consumers may not exactly be thick on the ground,Footnote ⁵ there are certainly artifacts – such as shear pins and fuses – that are designed to break. In due course we will consider whether cognitive systems have evolved any parallel.

3. Evolutionary design and dysfunction

Commercial disputes notwithstanding, the distinction between abnormal and normal functioning seems intuitive enough in the case of systems designed and manufactured by humans. How neatly, however, does this distinction carve nature at the joints? Is it equally clear for evolved, biological systems? In such cases, our criterion for determining malfunction (the disparity between actual functioning and intended functioning) would seem invalid, because (unlike the good people at Bolex) evolution is a blind watchmaker (Dawkins Reference Dawkins1986), without intentions. What we would like here is some way of making a distinction that is equivalent to the distinction between culpable design limitations and forgivable design limitations/features. Whereas culpable misdesign in manufactured items is the essence of artifactual malfunction, the evolutionary equivalent would be the marker of biological dysfunction.

Consider the human immune system. What would count as an example of immune system dysfunction? Presumably if the immune system were to succumb to a run-of-the-mill pathogen, we could speak uncontroversially of immune dysfunction. In some instances, however, the immune system “errs” in attempting to defend the body. Thus one of the main problems in organ transplants is that the immune system tries to protect the body against foreign matter, even a new heart that would ensure its survival. Is the activity of the immune system in the latter case strictly in accordance with its normal function? Perhaps that depends on what function we choose to impose upon the system. Insofar as the system functions to attack foreign matter, it has performed well. Insofar as the system is construed with the more general function of safeguarding the health of the body, however, it becomes less clear whether it has functioned normally – and this is the problem of evolutionary intentions-by-proxy.Footnote ⁶ Is all functionality just in the eye of the beholder? Millikan (Reference Millikan1984a; Reference Millikan1993) proposes a more objective solution to this problem:

According to Millikan, in order to determine the function of an organ or system we should consider not its present properties, powers, and dispositions, but instead its history.Footnote ⁷ Given that organ transplants have not featured in the evolutionary history of immune systems, any contemporary immune system that attacks a donor heart is functioning in accordance with the adaptive functioning of immune systems historically. That system, therefore, is functioning normally – or more precisely, Normally (see explanation below) – and its limitations are “forgivable.”

Let us consider a further parallel with our proposed misbelief taxonomy, this time by examining two types of misperception. Those of us who are short-sighted perceive (without our corrective lenses) a somewhat distorted visual world. Due to a kind of breakdown or degeneration, our visual systems (broadly construed) misrepresent the facts – they cease to function properly. Consider, on the other hand, what happens when we – with eyeglasses at the ready – submerge a perfectly straight stick into a pool of water. Do we continue to perceive the stick as straight and unbroken? No – our visual systems fail to compensate for the optical effect of refraction (they do not compute and correct for Snell's law; Boden Reference Boden and Hookway1984; Casperson Reference Casperson1999),Footnote ⁸ and the stick appears bent at the point where it meets the surface of the water. Our visual systems have again furnished us with misinformation, yet this time they have functioned Normally. The capital “N” here denotes a normative, rather than statistical, construal of “normal” (Millikan Reference Millikan1984a; Reference Millikan1993).

This is important because although our two examples of visual misperception (the short-sighted case and the stick-in-water case) can be distinguished on normative grounds (the first – being a case of visual dysfunction – is abNormal and the second Normal), they may both be “small-n” normal on statistical grounds. After all, the prevalence of myopia varies across ethnic groups, and is as high as 70–90% in some Asian populations (Chow et al. Reference Chow, Dhillon, Chew and Chew1990; Wong et al. Reference Wong, Foster, Hee, Ng, Tielsch, Chew, Johnson and Seah2000). Millikan (Reference Millikan1993), however, dismisses statistical construals of “normal” functioning. In a vivid example she points out that the proper function of sperm is to fertilise an ovum, notwithstanding the fact that, statistically speaking, it is exceedingly unlikely that any individual sperm will successfully perform that function (Millikan Reference Millikan1984a). Proper, Normal functioning, therefore, is not what happens always or even on the average; sometimes it is positively rare. Unless otherwise indicated, our subsequent usage of “normal” will follow Millikan's capitalised, normative sense.

Now, back to beliefs and misbeliefs. We contend that all instances of misbelief can be roughly classified as the output of either a dysfunctional, abnormal belief formation system or of a properly functioning, normal belief formation system. The former category, to which we turn briefly now, would not include adaptive misbeliefs (although see section 10), but provides a necessary background for understanding the better candidates – which, if they exist, will form a subset (design features) of the latter category.

4. Doxastic dysfunction

In the first category, misbeliefs result from breakdowns in the machinery of belief formation. If we conceive of the belief formation system as an information processing system that takes certain inputs (e.g., perceptual inputs) and (via manipulations of these inputs) produces certain outputs (beliefs, e.g., beliefs about the environment that the perceptual apparatus is directed upon), then these misbeliefs arise from dysfunction in the system – doxastic dysfunction. Such misbeliefs are the faulty output of a disordered, defective, abnormal cognitive system.

This view of misbelief is prominently exemplified by a branch of cognitive psychology known as cognitive neuropsychiatry (David & Halligan Reference David and Halligan1996). Cognitive neuropsychiatrists apply the logic of cognitive neuropsychology, which investigates disordered cognition in order to learn more about normal cognition, to disorders of high-level cognition such as delusions (Coltheart Reference Coltheart, Pashler and Wixted2002; Ellis & Young Reference Ellis and Young1988). Notwithstanding objections to the so-called doxastic conception of delusions (see sect. 9), delusions are misbeliefs par excellence – false beliefs that are held with strong conviction regardless of counterevidence and despite the efforts of others to dissuade the deluded individual (American Psychiatric Association 2000). They are first-rank symptoms of schizophrenia and prominent features of numerous other psychiatric and neurological conditions. Thematically speaking, delusions range from the bizarre and exotic (e.g., “I am the Emperor of Antarctica”; see David Reference David1999) to the more mundane and ordinary (e.g., “My husband is cheating on me”). Researchers in cognitive neuropsychiatry aim to develop a model of the processes involved in normal belief generation and evaluation, and to explain delusions in terms of damage to one or more of these processes.

To illustrate the cognitive neuropsychiatric approach to delusion, consider the case of “mirrored-self misidentification.” Patients with this rare delusion misidentify their own reflected image, and may come to believe that a stranger is following them around. Breen et al. (Reference Breen, Caine and Coltheart2001) investigated two cases of this delusion and uncovered two apparent routes to its development. The delusion of the first patient (“F.E.”) appeared to be underpinned by anomalous face perception (“prosopagnosia”), as he demonstrated a marked deficit in face processing on neuropsychological tests. In contrast, the face processing of the second patient (“T.H.”) was intact. This patient, however, appeared to be “mirror agnosic” (Ramachandran et al. Reference Ramachandran, Altschuler and Hillyer1997), in that he evinced an impaired appreciation of mirror spatial relations and was unable to interact appropriately with mirrors. His delusion appeared to be underpinned by anomalous processing not of faces, but of reflected space (see Breen et al. Reference Breen, Caine, Coltheart, Hendy and Roberts2000, for transcripts of interviews with the two patients; see Feinberg Reference Feinberg2001, Feinberg & Shapiro Reference Feinberg and Shapiro1989, and Spangenberg et al. Reference Spangenberg, Wagner and Bachman1998, for descriptions of related cases).

An important question arising at this point is the question of whether prosopagnosia (or mirror agnosia) is a sufficient condition for the mirror delusion. The answer to this question is almost certainly No. Other cases of mirror agnosia have been reported without any accompanying misidentification syndrome (Binkofski et al. Reference Binkofski, Buccino, Dohle, Seitz and Freund1999), and non-delusional prosopagnosia is quite common. Breen et al. (Reference Breen, Caine and Coltheart2001) thus proposed that the delusion of mirrored-self misidentification results from the conjunction of two cognitive deficits, the first of which gives rise to some anomalous perceptual data (data concerning either faces or reflected space), and the second of which allows the individual to accept a highly implausible hypothesis explaining these data. The first deficit accounts for the content of the delusion (the fact that it concerns a stranger in the mirror), while the second deficit accounts for why the stranger-in-the-mirror belief, once generated, is then adopted and maintained in the absence of appropriate evidence for that hypothesis. These deficits constitute breakdowns in the belief formation system, presumably underpinned by neuroanatomical or neurophysiological abnormalities. In both of the cases investigated by Breen et al. (Reference Breen, Caine and Coltheart2001), the mirror delusion occurred in the context of a progressive dementing illness.

Coltheart and colleagues (Coltheart et al., in press; Davies & Coltheart Reference Davies, Coltheart, Coltheart and Davies2000; Davies et al. Reference Davies, Coltheart, Langdon and Breen2001; Langdon & Coltheart Reference Langdon and Coltheart2000; McKay et al. Reference McKay, Langdon and Coltheart2007a; Reference McKay, Langdon, Coltheart, Bayne and Fernández2009) have suggested that a generalised framework of two concurrent cognitive deficits, or factors, might be used to explain delusions of many different types. In general, the first factor (Factor-1) consists of some endogenously generated abnormal data to which the individual is exposed. In addition to mirrored-self misidentification, Factors-1 have been identified or hypothesised that plausibly account for the content of delusions such as thought insertion, the Capgras delusion (the belief that a loved one has been replaced by an impostor) and the Cotard delusion (the belief that one is dead).

The second factor (Factor-2), on the other hand, can be characterised as a dysfunctional departure from Bayesian belief revision (Coltheart et al., in press), a departure that affects how beliefs are revised in the light of the abnormal Factor-1 data. Bayes' theorem is in a sense a prescription for navigating a course between excessive tendencies toward “observational adequacy” (whereby new data is over-accommodated) and “doxastic conservatism” (whereby existing beliefs are over-weighted) (Stone & Young Reference Stone and Young1997). McKay et al. (Reference McKay, Langdon, Coltheart, Bayne and Fernández2009) have suggested that whereas some delusions – for example, mirrored-self misidentification – might involve the former tendency (see Langdon & Coltheart Reference Langdon and Coltheart2000; Langdon et al. Reference Langdon, Cooper, Connaughton and Martin2006; Stone & Young Reference Stone and Young1997; also see Huq et al. Reference Huq, Garety and Hemsley1988), others – for example, delusional denial of paralysis (“anosognosia”) – might involve the latter (see Ramachandran Reference Ramachandran1994a; Reference Ramachandran, Sporns and Tononi1994b; Reference Ramachandran1995; Reference Ramachandran1996a; Reference Ramachandran1996b; Ramachandran & Blakeslee Reference Ramachandran and Blakeslee1998). In general, therefore, Factor-2 might be thought of as an acquired or congenital anomaly yielding one of two dysfunctional doxastic biases – a bias toward observational adequacy or toward doxastic conservatism.

The fact that we are not presently equipped with fail-safe belief-formation systems does not tell against an evolutionary perspective. This is because evolution does not necessarily produce optimally designed systems (Dawkins Reference Dawkins1982; Stich Reference Stich1990) and in fact often conspicuously fails to do so. It would be Panglossian to think otherwise (Gould & Lewontin Reference Gould and Lewontin1979; Voltaire Reference Voltaire and Smollett1759/1962):

Evolutionary explorations in Design Space are constrained, among other things, by economic considerations (beyond a certain level, system improvements may exhibit declining marginal utility; Stich Reference Stich1990), historical vicissitude (the appropriate mutations must occur if selection is to act on them), and the topography of the fitness landscape (selection cannot access optimal design solutions if it must traverse a fitness valley to do so; Dennett Reference Dennett1995a). Because evolution is an imperfect design process, the systems we have evolved for representing reality are bound to be limited – and sometimes they will break.

5. Misbeliefs as the output of a properly functioning system

Even if evolution were in some sense a “perfect” design process, there would still be limitations; only a violation of the laws of physics would permit, say, beliefs to be formed instantaneously, with no time lag whatsoever, or for an individual, finite believer to carry around in her head beliefs about the lack of prime factors of each specific prime number (only a brain of infinite volume could represent each individual prime).Footnote ⁹ The result is that even the beliefs of textbook Bayesians will frequently be false (or at least incomplete) – and such misbeliefs cannot be considered “culpable.”

Perhaps the most obvious examples of commonplace, forgivable misbelief occur when we are victimised by liars. Although extreme gullibility might be seen as dysfunctional (perhaps involving a Factor-2 bias toward observational adequacy), most of us (Bayesians included) are vulnerable to carefully crafted and disseminated falsehood. However adaptive it may be for us to believe truly, it may be adaptive for other parties if we believe falsely (Wallace Reference Wallace1973).Footnote ¹⁰ An evolutionary arms race of deceptive ploys and counterploys may thus ensue. In some cases the “other parties” in question may not even be animate agents, but cultural traits or systems (Dawkins Reference Dawkins2006a; Reference Dawkins2006b; Dennett Reference Dennett1995a; Reference Dennett2006a). Although such cases are interesting in their own right, the adaptive misbeliefs we pursue in this article are beneficial to their consumers – misbeliefs that evolve to the detriment of their believers are not our quarries.

So, given inevitable contexts of imperfect information, even lightning-fast Bayesians will frequently misbelieve, and such misbeliefs must be deemed forgivable. We briefly consider now whether certain departures from Bayesian updating might also be considered forgivable. Gigerenzer and colleagues (e.g., Gigerenzer & Goldstein Reference Gigerenzer and Goldstein1996; Gigerenzer et al. Reference Gigerenzer and Todd1999) have argued that some such departures, far from being defective, comprise “ecologically rational” decision strategies that operate effectively, given inevitable limitations of time and computational resources. These researchers have documented and investigated a series of such “fast and frugal” heuristics, including the “take the best” heuristic (Gigerenzer & Goldstein Reference Gigerenzer, Goldstein, Gigerenzer and Todd1999) and the “recognition heuristic” (Goldstein & Gigerenzer Reference Goldstein and Gigerenzer2002).

Some departures from normative rationality standards, however, result from perturbations in belief formation machinery and are not “heuristic” in any sense. As we have noted, bizarre misbeliefs like mirrored-self misidentification and the Cotard delusion may occur subsequent to neuropsychological damage. For example, Young et al. (Reference Young, Robertson, Hellawell, de Pauw and Pentland1992) described a patient who was injured in a serious motorcycle accident and subsequently became convinced that he was dead. Computerised tomography (CT) scans revealed contusions affecting temporo-parietal areas of this patient's right hemisphere as well as some bilateral damage to his frontal lobe. Misbeliefs, however, may also arise from less acute disruptions to the machinery of belief formation. For example, lapses in concentration due to fatigue or inebriation may result in individuals coming to hold erroneous beliefs, at least temporarily. Are such misbeliefs “culpable”? Do they reflect dysfunction in the belief formation system?

Although misbelief might always reflect the limitations of the system in some sense, it is not always easy to tell (absent a warranty) where imperfect proper doxastic functioning (forgivably limited) ends and where (culpably limited) doxastic dysfunction begins. This fuzziness is reflected in the literature on certain putative psychological disorders. As an example, consider the phenomenon of disordered reading. There are debates in the literature about whether there is a separate category of individuals who are disordered readers (e.g., see Coltheart Reference Coltheart1996). Opponents of this view argue that so-called “disordered readers” are just readers at the lower end of a Gaussian distribution of reading ability. Similarly, one of the most controversial psychiatric diagnoses in recent years has been the diagnosis of Attention-Deficit Hyperactivity Disorder (ADHD), which some commentators insist is a figment, arguing that putatively ADHD children are just children at the extreme ends of Gaussian distributions of attention and activity (for a discussion, see Dennett Reference Dennett, Conners and Kinsbourne1990a).

Controversies such as these are difficult to resolve. While we consider that Millikan's distinction between Normal and abNormal functioning provides a useful rule of thumb, we are not confident that this distinction – or any distinction – can be used to decisively settle disputes about forgivable versus culpable limitations in the biological domain. In this domain these categories are not discrete, but overlapping. Culpable misdesign in nature is always ephemeral – where design anomalies are rare or relatively benign, we will observe “tolerated” (forgivable) limitations; where anomalies begin to proliferate, however, they raise the selection pressure for a design revision, leading to either adaptive redesign or extinction. The upshot is that it may be difficult, if not impossible, to adjudicate on intermediate cases. How fatigued does an individual actually need to be before his doxastic lapses are deemed (evolutionarily) forgivable? And if alcohol did not feature in the evolutionary history of the belief formation system, are false beliefs formed while tipsy forgivable? Perhaps dousing one's brain in alcohol is akin to baking one's Bolex in a hot oven – both are forced to labour “under external conditions not Normal for performance of their proper functions” (Millikan Reference Millikan1993, p. 74; emphasis in original).

We acknowledge the overlap between our two broad categories of functioning. Such overlaps, however, characterise most biological categories: The boundaries – between, for example, species, or territories, or even between life and death – are porous and often violated. In any case, establishing a means of settling disputes about forgivable versus culpable limitations of the belief formation system is not crucial to our project. Although it is useful to be able to distinguish, crudely, between normal and abnormal doxastic functioning, the prevailing view is that misbeliefs formed in either case will themselves be abnormal. We will now begin to question this assumption. Contra the prevailing view, might there be certain situations in which misbelief can actually be adaptive (situations in which the misbeliefs themselves, not just the systems that produce them, are normal)? In those situations, if such there be, we would expect that we would be evolutionarily predisposed to form some misbeliefs. In short, misbelief would evolve.

6. Adaptive misbelief?

How could it ever be beneficial to believe a falsehood? Granted, one can easily imagine that in many circumstances it might feel better to misbelieve (more on this in sect. 10). Thus in Shakespeare's Richard II, Bolingbroke, who has been banished, is urged by his father to imagine that he is not banished but rather has left of his own volition. Bolingbroke's father appreciates that there may be psychological comfort in such a false belief. Bolingbroke's reply, however (“O! who can hold a fire in his hand…”), speaks both to the difficulty of deliberately misbelieving as well as to the apparent absence of tangible benefits in thus misbelieving. How could misbelief aid survival?

We note that it is easy to dream up anomalous offbeat scenarios where true beliefs are in fact detrimental for survival:

As Stich (Reference Stich1990) notes, cases such as this are highly unusual, and do little to refute the claim that true beliefs are generally adaptive (see also Millikan Reference Millikan1993). After all, natural selection does not act on anomalous particulars, but rather upon reliable generalizations. Our question, then, is whether there might be cases where misbelief is systematically adaptive.

7. The boy who cried wolf

Theoretical considerations converging from several different research traditions suggest that any such systematic falsehood must be unstable, yielding ephemeral instances, at best, of misbelief. Recognition of the problem is as old as Aesop's fable of the boy who cried wolf. Human communication between agents with memories and the capacity to check on the reliability of informants creates a dynamical situation in which systematic lying eventually exposes and discredits itself. As Quine (Reference Quine1960), Davidson (Reference Davidson and Tomberlin1994; Reference Davidson2001), Millikan (Reference Millikan2004), and other philosophers have noted, without a prevailing background of truth-telling, communication will erode, a practice that cannot pay for itself. That does not mean, of course, that individual liars will never succeed for long, but just that their success depends on their being rare and hard to track. A parallel phenomenon in evolutionary biology is Batesian mimicry, in which a non-poisonous species (or type within a species) mimics the appearance of a poisonous species (telling a falsehood about itself), getting protection against predators without paying for the venom. When mimics are rare, predators avoid them, having had more encounters with the poisonous variety; when mimics are common, the mimicry no longer works as well.

Quine and Ullian (Reference Quine and Ullian1978) note an important wrinkle:

Isolated cases of the tacit negation suggested in this passage actually occur, when what might be called systematic irony erodes itself with repetition. “Terrific” no longer means “provoking terror” but almost the opposite; and if somebody calls your lecture “incredible” and “fantastic,” you should not take offence – they almost certainly don't mean that they don't believe a word of it and deem it to be out of touch with reality. A related phenomenon is “grade inflation” in academia. “B+” just doesn't mean today what it used to mean several decades ago. When everybody is declared “better than average” the terms of the declaration are perforce diminished in meaning or credibility or both.

What, if anything, would prevent similar accommodations from diluting the effect of systematic falsehoods within the belief formation system of an individual organism? We know from many experiments with subjects wearing inverting or distorting lenses (for a recent summary, see Noë Reference Noë2004) that the falsehoods the eyes send the brain lead initially to false beliefs that seriously disable the subject, but in remarkably short time – a few days of accommodation – subjects have made an adjustment and can “get all the information from their testimony,” as Quine and Ullian (Reference Quine and Ullian1978) say, just as if they had inserted a tacit “not” or switched the meaning of “right” and “left” in the visual system's vocabulary. For a systematic falsehood-generating organ or tissue or network to have any staying power, it must send its lies to something that has little or no source memory or little or no plasticity in its evaluation of the credibility of the source.

Something like that may well be the case in some sensory systems. Akins (Reference Akins1996) discusses “narcissistic” biases built into sensory systems in order to optimize relevance and utility for the animal's behavioural needs. Instead of being designed to have their output states vary in unison (linearly) with the input conditions they are detecting (like thermometers or fuel gauges, which are designed to give objectively accurate measurements), these are designed to “distort” their responses (rather like the rear view mirror). She notes: “When a sensory system uses a narcissistic strategy to encode information, there need not be any counteracting system that has the task of decoding the output state” (p. 359). No “critics” or “lie detectors” devalue the message, and so the whole organism lives with a benign illusion of properties in the world that “just happen” to be tailor-made for its discernment. For instance, feedback from muscle stretch receptors needs to be discriminating over several orders of magnitude, so the “meaning” of the spike trains varies continuously over the range, the sensitivity being adjusted as need be to maintain fine-grained information over the whole range. “What is important to realize, here, is that there need not be any further device that records the ‘position’ of the gain mechanism.” (p. 362). In other words, no provision is made for reality-checking on what the stretch-receptors are “telling” the rest of the system, but the effect of this is to permit “inflation” to change the meaning of the spike frequency continuously.

Here, then, are two distinct ways in which our nervous systems can gracefully adjust the use to which they put signals that one would brand as false were it not for the adjustment. In the phenomena induced by artificially distorting the sensory input, we can observe the adjustment over time, with tell-tale behavioural errors and awkwardness giving way to quite effective and apparently effortless responses as the new meanings of the input signals get established. In the sort of cases Akins discusses, there is no precedent, no “traditional meaning,” to overcome, so there is no conflict to observe.

8. Alief and belief

Sometimes, however, the conflicts are not so readily resolved and the inconsistencies in behaviour do not evaporate. Gendler (Reference Gendler2008) notes the need for a category of quasi-beliefs and proposes to distinguish between alief and belief:

A person who trembles (or worse) when standing on the glass-floored Skywalk that protrudes over the Grand Canyon does not believe she is in danger, any more than a moviegoer at a horror film does, but her behaviour at the time indicates that she is in a belief-like state that has considerable behavioural impact. The reluctance of subjects in Paul Rozin's experiments with disgust (e.g., Rozin et al. Reference Rozin, Millman and Nemeroff1986) to come in contact with perfectly clean but disgusting looking objects, does not indicate that they actually believe the objects are contaminated; in Gendler's terms, they alieve this. In a similar vein, patients with Obsessive Compulsive Disorder (OCD) generally don't believe that the repetitive behaviours they feel compelled to engage in are necessary to prevent some dreaded occurrence – but they may well alieve this. (The Diagnostic and Statistical Manual of Mental Disorders [DSM-IV-TR; American Psychiatric Association 2000, p. 463] contains a specifier for OCD with “poor insight,” which denotes patients who fail to recognise that their obsessions and compulsions are “excessive or unreasonable.” In such patients alief may be overlaid with belief.)

Are such aliefs adaptive? Probably not. They seem to join other instances of “tolerated” side effects of imperfect systems, but in any case they are not beliefs proper. The question before us now is whether we ever evolve systems for engendering false beliefs: informational states of global and relatively enduring (inflation-proof) significance to the whole organism that miss the usual target of truth and do so non-coincidentally.

9. Error management theory

Beliefs are notoriously hard to count. Is the belief that 3+1 = 4 distinct from the belief that 1+3 = 4, or are these just one belief? Can you have one without the other? (See Dennett Reference Dennett and Woodfield1982, for an analysis of the problems attendant on such questions.) No matter how we individuate beliefs, we might expect that optimal systems of belief and decision would be maximally accurate. Given the contexts in which decisions are made, however, trade-offs may arise between overall accuracy and accuracy in certain situations. Dennett illustrates this point:

Stich echoes the logic of this scenario with an example of his own:

What these examples suggest is that when there are reliable “asymmetries in the costs of errors” (Bratman Reference Bratman1992) – that is, when one type of error (false positive or false negative) is consistently more detrimental to fitness than the other – then a system that is biased toward committing the less costly error may be more adaptive than an unbiased system. The suggestion that biologically engineered systems of decision and belief formation exploit such adaptations is the basis of Error Management Theory (EMT; Haselton Reference Haselton, Baumeister and Vohs2007; Haselton & Buss Reference Haselton and Buss2000; Reference Haselton, Buss, Forgas, Williams and Hippel2003; Haselton & Nettle Reference Haselton and Nettle2006). According to EMT, cognitive errors (including misbeliefs) are not necessarily malfunctions reflecting (culpable) limitations of evolutionary design; rather, such errors may reflect judicious systematic biases that maximise fitness despite increasing overall error rates.

Haselton and Buss (Reference Haselton and Buss2000) use EMT to explain the apparent tendency of men to overperceive the sexual interest and intent of women (Abbey Reference Abbey1982; Haselton Reference Haselton2003). They argue that, for men, the perception of sexual intent in women is a domain characterised by recurrent cost asymmetries, such that the cost of inferring sexual intent where none exists (a false-positive error) is outweighed by the cost of falsely inferring a lack of sexual intent (a false-negative). The former error may cost some time and effort spent in fruitless courtship, but the latter error will entail a missed sexual, and thus reproductive, opportunity – an altogether more serious outcome as far as fitness is concerned.

For women, the pattern of cost asymmetries is basically reversed. The cost of inferring a man's interest in familial investment where none exists (a false-positive error) would tend to outweigh the cost of falsely inferring a lack of such interest (a false-negative). The former error may entail the woman consenting to sex and being subsequently abandoned, a serious outcome indeed in arduous ancestral environments. The latter error, on the other hand, would tend merely to delay reproduction for the woman – a less costly error, especially given that reproductive opportunities are generally easier for women to acquire than for men (Haselton Reference Haselton, Baumeister and Vohs2007). In view of such considerations, proponents of EMT predict that women will tend to underperceive the commitment intentions of men, a prediction apparently supported by empirical evidence (Haselton Reference Haselton, Baumeister and Vohs2007; Haselton & Buss Reference Haselton and Buss2000).

Other EMT predictions that have received apparent empirical support include the hypotheses that recurrent cost asymmetries have produced evolved biases toward overinferring aggressive intentions in others (Duntley & Buss Reference Duntley and Buss1998; Haselton & Buss Reference Haselton and Buss2000), particularly in members of other racial and ethnic groups (Haselton & Nettle Reference Haselton and Nettle2006; Krebs & Denton Reference Krebs, Denton, Simpson and Kenrick1997; Quillian & Pager Reference Quillian and Pager2001); toward overinferring potential danger with regard to snakes (see Haselton & Buss Reference Haselton, Buss, Forgas, Williams and Hippel2003; Haselton & Nettle Reference Haselton and Nettle2006); toward underestimating the arrival time of approaching sound sources (Haselton & Nettle Reference Haselton and Nettle2006; Neuhoff Reference Neuhoff2001); and – reflecting Stich's (1990) example above – toward overestimating the likelihood that food is contaminated (see Rozin & Fallon Reference Rozin and Fallon1987; Rozin et al. Reference Rozin, Markwith and Ross1990). The error management perspective, moreover, appears to be a fecund source of new predictions. In the realm of sexuality and courtship, for example, Haselton and Nettle (Reference Haselton and Nettle2006) predict biases toward overinferring the romantic or sexual interest of (a) others in one's partner (what they term the “interloper effect”); and (b) one's partner in others. These predictions complement a series of other already confirmed predictions stemming from evolutionary analyses of jealousy (see Buss & Haselton [Reference Buss and Haselton2005] for a brief review).

One objection that might be raised at this point is that the above examples need not actually involve misbelief. Stich's omnivore need not believe that the food in question is poisonous – it might remain quite agnostic on that score. Similarly, jealous individuals need not harbour beliefs about partner infidelity – they might just be hypervigilant for any signs of it. The issue here is what doxastic inferences can be drawn from behaviour. After all, we always look before crossing a road, even where we are almost positive that there is no oncoming traffic. Our actions in such a case should not be read as reflecting a belief that there is an oncoming vehicle, but rather as reflecting a belief that there might be an oncoming vehicle (and the absence of a vehicle does not render that latter belief false). If we had to bet our lives one way or another on the matter, we might well bet that there isn't an oncoming vehicle (Bratman Reference Bratman1992). Betting our lives one way or the other, however, is a paradigm case of error symmetry (if we're wrong, we die – no matter which option we choose). In everyday cases of crossing the road, however, the errors are radically asymmetrical – an error one way may indeed mean serious injury or death, but an error the other way will entail only a trivial waste of time and energy.

The upshot of this criticism is that tendencies to “overestimate” the likelihood that food is contaminated, to “overperceive” the sexual interest of women, or to “overinfer” aggressive intentions in others, may reflect judicious decision criteria for action rather than misbeliefs. Nature may well prefer to create a bias on the side of prudence, but she does not always need to instill erroneous beliefs to accomplish this. She may instead make do with cautious action policies that might be expressed as “when in doubt [regarding some state of affairs relevant to current welfare], do x.” Errors, therefore, may not need to be managed doxastically. Some authors, however, have suggested that certain delusions also involve error management processes. Schipper et al. (Reference Schipper, Easton and Shackelford2007), for example, conceptualise delusional jealousy (also known as morbid jealousy or Othello syndrome) as the extreme end of a Gaussian distribution of jealousy, and hypothesise that the same sex-specific patterns that characterise “normal” jealousy – stemming from recurrent divergence in the adaptive problems faced by each gender – will also characterise delusional jealousy: “[H]ypersensitive jealousy mechanisms … may serve the adaptive purpose of preventing partner infidelity” (Schipper et al. Reference Schipper, Easton and Shackelford2007, p. 630; see also Easton et al. Reference Easton, Schipper and Shackelford2007). Whereas it may be true, therefore, that errors are not ordinarily managed doxastically, surely delusions involve genuine belief?

There are, however, serious objections to the notion that delusions are beliefs (Hamilton Reference Hamilton, Chung, Fulford and Graham2007; Stephens & Graham Reference Stephens and Graham2004; see Bayne & Pacherie [Reference Bayne and Pacherie2005] for a defence of the “doxastic conception”). One objection stems from the observation that although some individuals act on their delusions – and sometimes violently (see Mowat Reference Mowat1966; Silva et al. Reference Silva, Ferrari, Leong and Penny1998) – other deluded individuals frequently fail to act in accordance with their delusions. Individuals with the Capgras delusion, for example, rarely file missing persons reports on behalf of their replaced loved ones, and those who claim to be Napoleon are seldom seen issuing orders to their troops (Young Reference Young2000). In response to such objections, some authors have provided characterisations of delusions that dispense with the doxastic stipulation. Jaspers (1913/1963) and Berrios (Reference Berrios1991), for example, have each proposed “non-assertoric” accounts of delusions (Young Reference Young1999). Jaspers (1913/Reference Jaspers, Hoenig and Hamilton1963) held that schizophrenic delusions are not understandable, while for Berrios (Reference Berrios1991) the verbalizations of deluded patients are empty speech acts, mere noise masquerading as mentality. Other authors have put forward metacognitive accounts of delusions, whereby delusions are conceived as higher-order meta-evaluations of standard, lower-order mental items. For example, Currie and colleagues (Currie Reference Currie2000; Currie & Jureidini Reference Currie and Jureidini2001; see Bayne & Pacherie [2005] for a critique) argue that delusions are in fact imaginings misidentified as beliefs. On this account, the delusional belief of a Cotard patient is not the belief that she is dead, but rather the belief that she believes she is dead – when in fact she only imagines that she is dead (see Stephens & Graham [2004] for a variant of the metacognitive thesis).

In any case, it may be misguided to invoke delusions in attempting to link error management with adaptive misbelief. The reason is simple: Even if one overlooks objections to the doxastic conception and insists that delusions are beliefs, a serious problem remains – the issue of whether delusions can, in any sense, be regarded as adaptive. We consider this question below.

10. Doxastic shear pins

In this article we have distinguished two broad categories of misbelief – on the one hand, a category of misbeliefs resulting from breaks in the belief formation system, and on the other, a category of misbeliefs arising in the normal course of belief system operations. Here we briefly consider an intriguing intermediate possibility: misbeliefs enabled by the action of “doxastic shear pins.” A shear pin is a metal pin installed in, say, the drive train of a marine engine. The shear pin locks the propeller to the propeller shaft and is intended to “shear” should the propeller hit a log or other hard object. Shear pins are mechanical analogues of electrical fuses – each is a component in a system that is designed to break (in certain circumstances) so as to protect other, more expensive parts of the system. When a shear pin breaks (or a fuse blows), the system ceases its normal function. However, the action of the shear pin or fuse is not itself abnormal in these situations – in fact it is functioning perfectly as designed.

What might count as a doxastic analogue of shear pin breakage? We envision doxastic shear pins as components of belief evaluation machinery that are “designed” to break in situations of extreme psychological stress (analogous to the mechanical overload that breaks a shear pin or the power surge that blows a fuse). Perhaps the normal function (both normatively and statistically construed) of such components would be to constrain the influence of motivational processes on belief formation. Breakage of such components,Footnote ¹¹ therefore, might permit the formation and maintenance of comforting misbeliefs – beliefs that would ordinarily be rejected as ungrounded, but that would facilitate the negotiation of overwhelming circumstances (perhaps by enabling the management of powerful negative emotions) and that would thus be adaptive in such extraordinary circumstances.

Insofar as these misbeliefs were delusions, they would have a different aetiology to the more clear-cut cases of “deficit delusions” discussed earlier (mirrored-self misidentification and the like), because the breakage permitting their formation would serve a defensive, protective function. In short, they would be motivated (see Bayne & Fernández Reference Bayne, Fernández, Bayne and Fernández2009; McKay & Kinsbourne, in press; McKay et al. Reference McKay, Langdon and Coltheart2007a; Reference McKay, Langdon, Coltheart, Bayne and Fernández2009). Psychoanalytically inclined authors have proposed motivational interpretations of delusions such as the Capgras and Cotard delusions (e.g., see Enoch & Ball Reference Enoch and Ball2001), but in the wake of more rigorous cognitive neuropsychiatric models such interpretations tend to be viewed with disdain as outlandish and anachronistic (Ellis Reference Ellis2003).

Claims about motivational aetiologies for delusions are more plausible in other domains, however. Consider, for example, the following case of reverse Othello syndrome (Butler Reference Butler2000). The patient in question, “B.X.,” was a gifted musician who had been left a quadriplegic following a car accident. B.X. subsequently developed delusions about the continuing fidelity of his former romantic partner (who had in fact severed all contact with him and embarked on a new relationship soon after his accident). According to Butler, B.X.'s delusional system provided a “defense against depressive overwhelm … [going] some way toward reconferring a sense of meaning to his life experience and reintegrating his shattered sense of self. Without it there was only the stark reality of annihilating loss and confrontation with his own emotional devastation” (2000, p. 89). Although this seems a plausible motivational formulation, it comes from an isolated case study, and Butler's theorising is unavoidably post hoc. Moreover, the fact that B.X. had sustained severe head injuries in his accident opens up the possibility that any breakage in his belief evaluation system was, as it were, ateleological – adventitious, not designed. More general (plausible) motivational interpretations exist for other delusions, however – especially for so-called functional delusions, where the nature and role of underlying neuropathy (if any) is unspecified (Langdon & Coltheart Reference Langdon and Coltheart2000; Langdon et al. Reference Langdon, McKay, Coltheart, Freeman, Bentall and Garety2008). In particular, there are well-worked-out motivational formulations for persecutory delusions (see Bentall & Kaney Reference Bentall and Kaney1996; Kinderman & Bentall Reference Kinderman and Bentall1996; Reference Kinderman and Bentall1997), interpretations that have garnered recent empirical support (McKay et al. Reference McKay, Langdon and Coltheart2007b; Moritz et al. Reference Moritz, Werner and von Collani2006; although, see Vazquez et al. Reference Vazquez, Diez-Alegria, Hernandez-Lloreda and Moreno2008).

It seems, therefore, that certain delusions might serve plausible defensive functions. Whether this implies that such delusions are adaptive, however, is a different question. To be sure, it might plausibly be argued that delusions are psychologically adaptive in certain scenarios (as the above reverse Othello case suggests). But this does not establish a case for biological adaptation. Here we must be careful to honour a distinction, often complacently ignored, between human happiness and genetic fitness. If the most promising path, on average, to having more surviving grandoffspring is one that involves pain and hardship, natural selection will not be deterred in the least from pursuing it (it is well to remind ourselves of the insect species in which the males are beheaded in the normal course of copulation, or – somewhat closer to home – the ruthless siblingcide practiced by many bird species). Perhaps the most that can presently be claimed is that delusions may be produced by extreme versions of systems that have evolved in accordance with error management principles, that is, evolved so as to exploit recurrent cost asymmetries. As extreme versions, however, there is every chance that such systems manage errors in a maladaptive fashion. As Zolotova and Brüne conclude, “[T]he content of delusional beliefs could be interpreted as pathological variants of adaptive psychological mechanisms…” (Reference Zolotova and Brüne2006, p. 192, our emphasis; see also Brüne Reference Brüne2001; Reference Brüne and Shohov2003a; Reference Brüne2003b).

In view of these caveats, it is unclear whether delusions could form via the teleological “shearing” of particular belief components under stressful circumstances. Non-delusional misbeliefs, however, might potentially be formed in something like this way (see section 13 for a discussion of health illusions). To an extent the issue here is merely stipulative, hinging on the definition of “delusion” one adopts. If delusions are dysfunctional by definition, then they cannot be adaptive. Moreover, many have reported that, in times of great stress, faith in God has given them “the strength to go on.” It may be true that there are no atheists in foxholes (although see Dennett Reference Dennett2006b), but if delusions are defined so as to exclude conventional religious beliefs (American Psychiatric Association 2000), then even if foxhole theism is biologically adaptive it will not count as an instance of biologically adaptive delusion.

Accounts of religious belief as an adaptation in general have been proposed by a number of commentators (e.g., Johnson & Bering Reference Johnson and Bering2006; Wilson Reference Wilson2002; but see Dennett [2006a] for a critique and an alternative evolutionary account). Given the costs associated with religious commitment (see Bulbulia Reference Bulbulia2004b; Dawkins Reference Dawkins2006a; Ruffle & Sosis Reference Ruffle and Sosis2007; Sosis Reference Sosis2004), it seems likely that such commitment is accompanied by bona fide belief of one sort or another (it might be only bona fide belief in belief – see Dennett Reference Dennett2006a). We therefore consider now whether in religion we have a candidate domain of adaptive misbelief.

11. Supernatural agency

Interestingly, error management logic pervades contemporary thinking about the origin of religion, and it is also apparent in some less-contemporary thinking:

Pascal's famous wager provides perhaps the quintessential statement of error management logic, although it is important to note that the wager is an outcome of domain general rationality, whereas error management as implemented by evolved cognitive mechanisms is always domain specific (Haselton & Nettle Reference Haselton and Nettle2006). One such domain relevant to religion is the domain of agency detection. Guthrie (Reference Guthrie1993) has argued that a bias toward inferring the presence of agents would have been adaptive in the evolutionary past: “It is better for a hiker to mistake a boulder for a bear, than to mistake a bear for a boulder” (1993, p. 6). He argues further that religious belief may be a by-product of evolved cognitive mechanisms that produce such biases – mechanisms that Barrett (Reference Barrett2000) has termed “Hyperactive agent-detection devices” (HADDs). As a by-product theory of religion (see further on), this account provides little suggestion that religious belief is adaptive misbelief. Other authors, however, have proposed accounts of religion as an adaptation that incorporate error management logic.

For example, Johnson, Bering, and colleagues (Bering & Johnson Reference Bering and Johnson2005; Johnson Reference Johnson2005; Johnson & Bering Reference Johnson and Bering2006; Johnson & Krüger Reference Johnson and Krüger2004; Johnson et al. Reference Johnson, Stopka and Knights2003) have advanced a “supernatural punishment hypothesis” regarding the evolution of human cooperation. The nature and extent of human cooperation poses a significant evolutionary puzzle (Fehr & Gaechter Reference Fehr and Gaechter2002). Human societies are strikingly anomalous in this respect relative to other animal species, as they are based on large-scale cooperation between genetically unrelated individuals (Fehr & Fischbacher Reference Fehr and Fischbacher2003; Reference Fehr and Fischbacher2004). Classic adaptationist accounts of cooperation such as kin selection (Hamilton Reference Hamilton1964) and direct reciprocity (Trivers Reference Trivers1971) cannot explain these features of human cooperation. Moreover, the theories of indirect reciprocity (Alexander Reference Alexander1987) and costly signalling (Gintis et al. Reference Gintis, Smith and Bowles2001; Zahavi Reference Zahavi1995), which show how cooperation can emerge in larger groups when individuals have the opportunity to establish reputations, struggle to explain the occurrence of cooperation in situations that preclude reputation formation – such as in anonymous, one-shot economic games (Fehr & Gaechter Reference Fehr and Gaechter2002; Gintis et al. Reference Gintis, Bowles, Boyd and Fehr2003; Henrich & Fehr Reference Henrich, Fehr and Hammerstein2003).

Johnson, Bering, and colleagues (Bering & Johnson Reference Bering and Johnson2005; Johnson Reference Johnson2005; Johnson & Bering Reference Johnson and Bering2006; Johnson & Krüger Reference Johnson and Krüger2004; Johnson et al. Reference Johnson, Stopka and Knights2003) argue that belief in morally interested supernatural agents – and fear of punishment by such agents – may sustain cooperation in such situations. The argument they put forward is based explicitly on error management theory. They suggest that the evolutionary advent of language, on the one hand, and Theory of Mind (ToM; Premack & Woodruff Reference Premack and Woodruff1978), on the other (specifically, the evolution of the “intentionality system,” a component of ToM geared toward representing mental states as the unseen causes of behaviour; Bering Reference Bering2002; Povinelli & Bering Reference Povinelli and Bering2002), occasioned a novel set of selection pressures. In particular, the evolution of these cognitive capabilities increased the costs associated with social defection (because one's social transgressions could be reported to absent third parties), and thus increased the adaptiveness of mechanisms that inhibit selfish actions.

Belief in supernatural punishment – an incidental by-product of the intentionality system – is one such mechanism. Johnson, Bering, and colleagues thus argue for supernatural belief as an exaptation (Gould & Vrba Reference Gould and Vrba1982), a fact that is important for the plausibility of their model. Their central claim is that selection would favour exaggerated estimates of the probability and/or consequences of detection, and thus would favour belief in morally interested supernatural agents. It is not clear, however, that the latter would be necessary to drive the former. Selection might simply implement biased beliefs regarding the probability and/or consequences of detection (cutting out the middleman, as it were). Even more parsimoniously, selection might favour accurate beliefs and implement appropriately judicious action policies vis-à-vis social situations (cf. the social exchange heuristic of Yamagishi et al. Reference Yamagishi, Terai, Kiyonari, Mifune and Kanazawa2007). As per our earlier observations regarding evolutionary explorations in Design Space, however, such “simpler” solutions might be unavailable to selection; it may be that the most direct means of inhibiting selfish behaviour is via supernatural punishment beliefs. If such beliefs were already on the evolutionary scene as by-products of pre-existing intentionality system structures, then they could be conveniently co-opted without any need for the engineering of novel neuro-cognitive machinery (see Bering Reference Bering2006).

The argument depends on a crucial error management assumption – that the costs of the two relevant errors in this novel selection environment are recurrently asymmetric – that is, that the cost of cheating and being caught reliably exceeds the cost of cooperating when cheating would have gone undetected. Provided that this inequality obtains, the theory claims that a propensity to believe in morally interested supernatural agents would have been selected for, because individuals holding such beliefs would tend to err on the (cooperative) side of caution in their dealings with conspecifics. “Machiavellian” unbelievers would not therefore gain an advantage, as they would lack important “restraints on self-interested conduct” and thus be “too blatantly selfish for the subtleties of the new social world” (Johnson Reference Johnson2005, p. 414).

What is the evidence for this theory? Johnson (Reference Johnson2005) utilized data from Murdock and White's (Reference Murdock and White1969) Standard Cross-Cultural Sample (SCCS) of 186 human societies around the globe to test whether the concept of supernatural punishment – indexed by the importance of moralizing “high gods” – was associated with cooperation. Johnson found “high gods” to be “significantly associated with societies that are larger, more norm compliant in some tests (but not others), loan and use abstract money, are centrally sanctioned, policed, and pay taxes” (Johnson Reference Johnson2005, p. 426; see also Roes & Raymond Reference Roes and Raymond2003). As Johnson acknowledges, his measures of supernatural punishment and cooperation were imprecise (a limitation of the data set employed), and his evidence is correlational at best – the causal relationship between supernatural punishment beliefs and cooperation remains obscure. The same criticisms apply to Rossano's (Reference Rossano2007) argument that the emergence (in the Upper Palaeolithic) of certain ancient traits of religion (involving belief in “ever-vigilant spiritual monitors”; p. 272) coincides with evidence for a dramatic advance in human cooperation (see Norenzayan & Shariff [Reference Norenzayan and Shariff2008] for a review of further studies reporting correlational evidence of religious prosociality).

In view of this criticism, studies that elicit causal evidence for the supernatural punishment hypothesis are crucial. The findings of a recent study by Shariff and Norenzayan (Reference Shariff and Norenzayan2007) are worth considering in this regard. These authors used a scrambled-sentence paradigm to implicitly prime “God” concepts, and found that participants primed in this manner gave significantly more money in a subsequent (anonymous, one-shot) economic game (the Dictator Game; see Camerer Reference Camerer2003) than control participants. In discussing these results, Shariff and Norenzayan made appeal to a “supernatural watcher” interpretation of their findings, suggesting that their religious primes “aroused an imagined presence of supernatural watchers, and that this perception then increased prosocial behavior” (p. 807). As Randolph-Seng and Nielsen (Reference Randolph-Seng and Nielsen2008) note, however, this interpretation may be less parsimonious than a behavioural-priming or ideomotor-action account (which Shariff and Norenzayan also considered), in which the activation of specific perceptual-conceptual representations increases the likelihood of behaviour consistent with those representations (see Dijksterhuis et al. Reference Dijksterhuis, Chartrand, Aarts and Bargh2007). Thus, much as people walk more slowly when the concept “elderly” is primed (Bargh et al. Reference Bargh, Chen and Burrows1996), priming words that are semantically associated with prosocial behaviour (including words such as “God” and “prophet,” both of which were utilised as “religious primes” by Shariff and Norenzayan) may lead to such behaviour simply by virtue of that association.

The behavioural-priming or ideomotor-action explanation is buttressed by the results of Shariff and Norenzayan's second study, which showed that implicitly primed “secular” concepts were comparable to implicitly primed “God” concepts in terms of their effect on giving in a subsequent Dictator Game. As Randolph-Seng and Nielsen (Reference Randolph-Seng and Nielsen2008) point out, it is not clear why secular primes such as “civic” and “contract,” that contain no reference to God, should enhance prosocial behaviour if such behaviour results from the activation of “supernatural watcher” concepts. Nevertheless, we feel that the research design of Shariff and Norenzayan (and that of comparable recent studies; see Pichon et al. Reference Pichon, Boccato and Saroglou2007; Randolph-Seng & Nielsen Reference Randolph-Seng and Nielsen2007) is insufficient to adequately discriminate between the supernatural watcher and behavioural-priming interpretations. What is needed is a study that clearly separates the influence of an “agency” dimension (whether natural or supernatural) from a “prosociality” dimension. The appeal of the supernatural punishment hypothesis is that it shows how reputational concerns might influence behaviour in situations that preclude actual reputation formation. It is true that both the “religious prime” and the “secular prime” categories utilized by Shariff and Norenzayan included words potentially associated semantically with prosocial behaviour. We note, however, that both word categories also include words potentially associated with agency (“God” and “prophet” in the former category, “jury” and “police” in the latter). It may be that the surveillance connotations of a word such as “police” may mean that priming with this word enhances prosocial behaviour by activating reputational concerns – not by semantic association with prosociality! Future studies would do well to tease these factors apart.

Recent research by Bering et al. (Reference Bering, McLeod and Shackelford2005) employed a different paradigm to elicit causal evidence regarding the effect of a supernatural watcher (albeit a supernatural watcher without obvious moral interests). In one condition of their third study, undergraduate students were casually informed that the ghost of a dead graduate student had recently been noticed in the testing room. These participants were subsequently less willing than control participants to cheat on a competitive computer task, despite a low apparent risk of social detection. This result is intriguing, and not obviously susceptible to explanation in terms of behavioural-priming effects (cf. Randolph-Seng & Nielsen Reference Randolph-Seng and Nielsen2007). As the relevant information was not collected, however, it is not clear to what extent the effect of the ghost prime in this study was mediated by participants' belief in ghosts. This is an important point, as it raises the possibility that if behavioural effects are reliably elicited by supernatural primes, they may be elicited not by belief but by alief(!) (Gendler Reference Gendler2008). Perhaps suitably primed participants alieve that a supernatural agent is watching, but believe no such thing. If this is the case, then such effects, although interesting, will have little bearing on the question of whether misbelief can be systematically adaptive.

It turns out that the evidence is mixed regarding whether supernatural belief mediates the effect of supernatural primes on behaviour. In the first of Shariff and Norenzayan's (2007) studies, the religious prime increased generosity for both theists and atheists. In their second study, however, the effect of the religious prime was stronger for theists than atheists (and in fact non-significant for atheists). It may be that this difference is attributable to the more stringent atheist criterion employed in the latter study, in which case belief may be crucial. Recent work by Bushman et al. (Reference Bushman, Ridge, Das, Key and Busath2007), which found that scriptural violence sanctioned by God increased aggression, especially in religious participants, is consistent with this proposition. However, Randolph-Seng and Nielsen (Reference Randolph-Seng and Nielsen2007) found that whereas participants primed with religious words cheated significantly less on a subsequent task than control participants, the intrinsic religiosity of participants did not interact with the prime factor.

At present, therefore, there is no strong evidence that religious belief is important for the efficacy of religious primes, nor any strong evidence that such primes exert their effects by activating reputational concerns involving supernatural agents. Other approaches notwithstanding (e.g., Dawkins Reference Dawkins2006a; Sosis Reference Sosis2004; Wilson Reference Wilson2002), the currently dominant evolutionary perspective on religion remains a by-product perspective (Atran Reference Atran2004; Atran & Norenzayan Reference Atran and Norenzayan2004; Bloom Reference Bloom2004; Reference Bloom2005; Reference Bloom2007; Boyer Reference Boyer2001; Reference Boyer2003; Reference Boyer2008b; Hinde Reference Hinde1999). On this view, supernatural (mis)beliefs are side-effects of a suite of cognitive mechanisms adapted for other purposes. Such mechanisms render us hyperactive agency detectors (Barrett Reference Barrett2000; Guthrie Reference Guthrie1993), promiscuous teleologists (Kelemen Reference Kelemen2004), and intuitive dualists (Bloom Reference Bloom2004); collectively (and incidentally), they predispose us to develop religious beliefs – or at least they facilitate the acquisition of such beliefs (Bloom Reference Bloom2007). Meanwhile, advocates of “strong reciprocity” (Fehr et al. Reference Fehr, Fischbacher and Gaechter2002; Gintis Reference Gintis2000) argue that the puzzle of large-scale human cooperation may be solved by invoking cultural group selection (Boyd et al. Reference Boyd, Gintis, Bowles and Richerson2003; Henrich & Boyd Reference Henrich and Boyd2001) or gene-culture coevolution (Bowles et al. Reference Bowles, Choi and Hopfensitz2003; also see Fehr & Fischbacher Reference Fehr and Fischbacher2003; Gintis Reference Gintis2003).

12. Self-deception

Arguments that systematic misbelief may have been selected for its ability to facilitate the successful negotiation of social exchange scenarios are not confined to the domain of religion. In his foreword to the first edition of Richard Dawkins' book The Selfish Gene, for example, the evolutionary biologist Robert Trivers outlined an influential theory of the evolution of self-deception:

In the intervening years the notion that self-deception has evolved because it facilitates other-deception appears to have become something of a received view in evolutionary circles. The notion is not without its critics, however. Both Ramachandran and Blakeslee (Reference Ramachandran and Blakeslee1998) and Van Leeuwen (Reference Van Leeuwen2007) have pointed out that deceivers who believe their own lies (regarding, say, the whereabouts of a food source) will not themselves be able to take advantage of the truth. Deception is thus clearly possible without self-deception. Van Leeuwen (Reference Van Leeuwen2007) also claims the converse – that self-deception frequently occurs in the absence of any intention to deceive. On the basis of such considerations, Van Leeuwen argues that self-deception is not an adaptation but a by-product of other features of human cognitive architecture.

In any case, Trivers' theory has received surprisingly little empirical attention, and we know of no direct empirical evidence that the theory is valid. Indeed, a recent study by McKay et al. (in preparation) found preliminary evidence that high self-deceivers were, if anything, less likely to be trusted in a cooperative exchange situation than low self-deceivers. These authors recruited groups of previously unacquainted participants, had them interact briefly with one another, and then invited each participant to play an anonymous, one-shot Prisoner's Dilemma game with each other participant. Participants were subsequently told that they could double the stakes for one of these games. Individuals higher in self-deception (measured using the Self-Deceptive Enhancement [SDE] scale of the Balanced Inventory of Desirable Responding [BIDR; Paulhus Reference Paulhus1988]) were less likely to be nominated for such double-stakes exchanges, suggesting that such individuals appeared less trustworthy than individuals lower in self-deception.

In a variant of Trivers' dictum, Krebs and Denton (Reference Krebs, Denton, Simpson and Kenrick1997) state that “Illusions about one's worth are adaptive because they help people deceive others about their worth” (p. 37; see also Smith Reference Smith2006). Given the lack of evidence that others are deceived about the worth of self-deceptive individuals, it is questionable whether “illusions about one's worth” do in fact serve this function. Might such illusions serve other adaptive functions, however? Having peeled the onion down, and set aside a variety of inconclusive candidates for adaptive misbelief, we turn finally to an investigation of this question.

13. Positive illusions

In parallel with the prevailing evolutionary view of adaptive belief, a number of psychological traditions have regarded close contact with reality as a cornerstone of mental health (Jahoda Reference Jahoda1953; Reference Jahoda1958; Maslow Reference Maslow1950; Peck Reference Peck1978; Vaillant Reference Vaillant1977). A substantial body of research in recent decades, however, has challenged this view, suggesting instead that optimal mental health is associated with unrealistically positive self-appraisals and beliefs.Footnote ¹² Taylor and colleagues (e.g., Taylor Reference Taylor1989; Taylor & Brown Reference Taylor and Brown1988) refer to such biased perceptions as “positive illusions,” where an illusion is “a belief that departs from reality” (Taylor & Brown Reference Taylor and Brown1988, p. 194). Such illusions include unrealistically positive self-evaluations, exaggerated perceptions of personal control or mastery, and unrealistic optimism about the future.

For example, evidence indicates that there is a widespread tendency for most people to see themselves as better than most others on a range of dimensions. This is the “better-than-average effect” (Alicke Reference Alicke1985) – individuals, on the average, judge themselves to be more intelligent, honest, persistent, original, friendly, and reliable than the average person. Most college students tend to believe that they will have a longer-than-average lifespan, while most college instructors believe that they are better-than-average teachers (Cross Reference Cross1977). Most people also tend to believe that their driving skills are better than average – even those who have been hospitalised for accidents (see, e.g., McKenna et al. Reference McKenna, Stanier and Lewis1991; Williams Reference Williams2003). In fact, most people view themselves as better than average on almost any dimension that is both subjective and socially desirable (Myers Reference Myers2002). Indeed, with exquisite irony, most people even see themselves as less prone to such self-serving distortions than others (Friedrich Reference Friedrich1996; Pronin et al. Reference Pronin, Lin and Ross2002; Pronin Reference Pronin, Gilovich and Ross2004).

Positive illusions may well be pervasive, but are they adaptive, evolutionarily speaking? For example, do such misbeliefs sustain and enhance physical health? Our positive illusions may “feel good” and yet contribute nothing to – or even be a tolerable burden upon – our genetic fitness, a side effect that evolution has not found worth blocking. On the other hand, they may be fitness-enhancing, in either of two quite different ways. They may lead us to undertake adaptive actions; or they may more directly sustain and enhance health, or physical fitness in the everyday sense. We consider each of these prospects in turn.

First, let's look at what happens when positive illusions affect the decisions we make in the course of deliberate, intentional action. Do these rosy visions actually lead people to engage in more adaptive behaviours? According to Taylor and Brown (Reference Taylor and Brown1994b), they do. These authors note that individuals with strong positive perceptions – and in particular, inflated perceptions – of their abilities are more likely to attain success than those with more modest self-perceptions. In this connection they quote Bandura:

Haselton and Nettle (Reference Haselton and Nettle2006) note the tacit error management perspective in Taylor and Brown's conception of positive illusions:

Although the link here with error management is interesting and relevant, it is worth pausing to consider the precise wording of this quote. Haselton and Nettle speak of an illusional positive belief as being better than an unbiased belief, when presumably what they mean is that a belief system geared toward forming illusional positive beliefs – assuming that such beliefs are consistently less detrimental to fitness than illusional negative beliefs – may be more adaptive than an unbiased belief system. Even if the misbeliefs arising through the operation of the former system arise through the normal operation of that system, the misbeliefs themselves must surely count as abnormal (Millikan Reference Millikan2004). After all, it's not clear that there is anything adaptive about trying and failing (but see Dennett Reference Dennett, Brockman and Matson1995b). Smoke detectors biased toward false alarms are no doubt preferable to those biased toward the more costly errors (failures to detect actual fires); but that doesn't mean that a false alarm is a cause for celebration. If a smoke detector came onto the market that detected every actual fire without ever sounding a false alarm, that would be the one to purchase. Even if they spring from adaptively biased misbelief-producing systems, therefore, individual misbeliefs about success are arguably more of a tolerable by-product than an adaptation. (Possible exceptions to this might be cases where individuals falsely believe that they will attain great success, yet where the confident striving engendered by such misbelief leads to greater success than would have been attained had they not falsely believed. Perhaps it is sometimes necessary to believe that you will win gold in order to have any chance of winning silver or bronze; see Benabou & Tirole Reference Benabou and Tirole2002; Krebs & Denton Reference Krebs, Denton, Simpson and Kenrick1997).

Might there be evidence, however, that misbeliefs themselves can propel adaptive actions? Here we note that positive illusions need not be merely about oneself. Perhaps the most compelling indication that positively biased beliefs lead people to engage in biologically adaptive behaviours is when such beliefs concern other people – in particular, those we love. Gagné and Lydon (Reference Gagné and Lydon2004; see also Fowers et al. Reference Fowers, Lyons and Montel1996; Fowers et al. Reference Fowers, Lyons, Montel and Shaked2001; Murray et al. Reference Murray, Holmes and Griffin1996) have found that the better-than-average effect applies for people's appraisals not just of themselves but also of their partners: 95% judge their partners more positively than the average partner with respect to intelligence, attractiveness, warmth, and sense of humour. Such biased appraisal mechanisms may be crucial to ensure the completion of species-specific parental duties: “The primary function of love is to cement sexual relationships for a period of several years, in order to ensure that the vulnerable human infant receives care from its mother, resources from its father, and protection from both” (Tallis Reference Tallis2005, p. 194; see also Fisher Reference Fisher, Sternberg and Weis2006). Note, in this connection, that biased appraisals of one's children may also facilitate parental care: “[T]he ability of parents to deny the faults of their children sometimes seems to border on delusion” (Krebs & Denton Reference Krebs, Denton, Simpson and Kenrick1997, p. 34). Wenger and Fowers (Reference Wenger and Fowers2008) have recently provided systematic evidence of positive illusions in parenting. Most participants in their study rated their own children as possessing more positive (86%) and fewer negative (82%) attributes than the average child. This better-than-average effect, moreover, was a significant predictor of general parenting satisfaction.

Finally, we consider evidence that positive illusions can directly sustain and enhance health. Research has indicated that unrealistically positive views of one's medical condition and of one's ability to influence it are associated with increased health and longevity (Taylor et al. Reference Taylor, Lerner, Sherman, Sage and McDowell2003). For example, in studies with HIV-positive and AIDS patients, those with unrealistically positive views of their likely course of illness showed a slower illness course (Reed et al. Reference Reed, Kemeny, Taylor and Visscher1999) and a longer survival time (Reed et al. Reference Reed, Kemeny, Taylor, Wang and Visscher1994; for a review, see Taylor et al. Reference Taylor, Kemeny, Reed, Bower and Gruenewald2000).

Taylor et al. (Reference Taylor, Kemeny, Reed, Bower and Gruenewald2000) conjectured that positive illusions might work their medical magic by regulating physiological and neuroendocrine responses to stressful circumstances. Stress-induced activation of the autonomic nervous system and the hypothalamic-pituitary-adrenocortical (HPA) axis facilitates “fight or flight” responses and is thus adaptive in the short-term. Chronic or recurrent activation of these systems, however, may be detrimental to health (see McEwen Reference McEwen1998), so psychological mechanisms that constrain the activation of such systems (perhaps doxastic shear pins that break – or even just bend a little – in situations of heightened stress) may be beneficial. Consistent with the above hypothesis, Taylor et al. (Reference Taylor, Lerner, Sherman, Sage and McDowell2003) found that self-enhancing cognitions in healthy adults were associated with lower cardiovascular responses to stress, more rapid cardiovascular recovery, and lower baseline cortisol levels.

Results linking positive illusions to health benefits are consistent with earlier findings that patients who deny the risks of imminent surgery suffer fewer medical complications and are discharged more quickly than other patients (Goleman Reference Goleman1987), and that women who cope with breast cancer by employing a denial strategy are more likely to remain recurrence-free than those utilising other coping strategies (Dean & Surtees Reference Dean and Surtees1989). In such cases the expectation of recovery appears to facilitate recovery itself, even if that expectation is unrealistic. This dynamic may be at work in cases of the ubiquitous placebo effect, whereby the administration of a medical intervention instigates recovery before the treatment could have had any direct effect and even when the intervention itself is completely bogus (Benedetti et al. Reference Benedetti, Pollo, Lopiano, Lanotte, Vighetti and Rainero2003; Humphrey Reference Humphrey and Gregory2004).

Placebos have been acclaimed, ironically, as “the most adaptable, protean, effective, safe and cheap drugs in the world's pharmacopoeia” (Buckman & Sabbagh Reference Buckman and Sabbagh1993, p. 246). They have proven effective in the treatment of pain and inflammation, stomach ulcers, angina, heart disease, cancer, and depression, among other conditions (Humphrey Reference Humphrey2002; Reference Humphrey and Gregory2004). From an evolutionary perspective, however, the placebo effect presents something of a paradox:

Humphrey (Reference Humphrey2002; 2004) considers the placebo effect in an evolutionary context and suggests an ingenious solution to this paradox. Noting that immune system functioning can be very costly, Humphrey construes the human immune response as under the regulation of an evolved administrative system that must manage resources as efficiently as possible. Because resources are limited, there is adaptive value to limiting resource expenditure just as there is value in the expenditure itself.

Sound economic management requires forecasting the future, and thus the health management system would need to take into account any available information relevant to future prospects. Such data would include information about the nature of the threat itself (including the likelihood of spontaneous remission), the costs of mounting an appropriate defence, and evidence relating to the course of the illness in other victims. Paramount among such sources of information, however, would be information about the availability of medical care: “People have learned … that nothing is a better predictor of how things will turn out when they are sick … than the presence of doctors, medicines, and so on” (Humphrey Reference Humphrey and Gregory2004, p. 736). To put a military gloss on Humphrey's economic resource management metaphor, there is less need for caution and conservation of resources once reinforcements arrive. Only then can one “spare no expense in hopes of a quick cure” (Dennett Reference Dennett2006a, p. 138, emphasis in original).

The placebo effect seems at first to be a case where misbelief in the efficacy of a particular treatment regimen (which, after all, may be a sham with zero direct efficacy) facilitates health and physical fitness. Is this, however, a case of evolved misbelief? If Humphrey's account of the placebo effect is along the right lines, what evolved was a bias to attend to and wait for signs of security before triggering a full-bore immune response, and these signs would, in the main, have been true harbingers of security (otherwise the bias would not have been adaptive and would not have evolved). As drug trials and placebos did not figure in our evolutionary history, they represent a later, artificial “tricking” of this evolved system, similar to the way calorie-free saccharine tricks our sweet tooth or pornography tricks our libido. Placebo misbelief, therefore, is not adaptive misbelief – it is a by-product of an adaptation. In Humphrey's words, the “human capacity for responding to placebos is … an emergent property of something else that is genuinely adaptive: namely, a specially designed procedure for ‘economic resource management’…. Unjustified placebo responses, triggered by invalid hopes, must be counted a biological mistake” (Humphrey Reference Humphrey2002, pp. 261 and 279, emphasis in original).

Do similar remarks apply to the instances of positive illusion and health discussed above? Are the “unjustified” expectations and “invalid hopes” of some AIDS and cancer patients biologically mistaken? One might argue that if “unrealistic” optimism facilitates happy outcomes, then – in retrospect – such optimism was not so unrealistic after all! However, it seems clear that optimism in the relevant studies is not realistic optimism (even allowing that this is not an oxymoronic concept). For example, Reed et al. (Reference Reed, Kemeny, Taylor, Wang and Visscher1994) recruited gay men, who had been diagnosed with AIDS for about a year, for an investigation into the effect of positive illusions on physical health. As the data for this particular study were collected in the late 1980s, life expectancy for these men was not long, and two-thirds of the men had died by the time of completion of the study. Realistic acceptance of death (measured by items including the reverse-scored item “I refuse to believe that this problem has happened”) was found to be a significant negative predictor of longevity, with high scorers on this measure typically dying nine months earlier than low scorers. This relationship remained significant when a variety of potential predictors of death were controlled for, including age, time since diagnosis, self-reported health status, and number of AIDS-related symptoms. It does seem, therefore, that the relevant beliefs here were unrealistically positive. “Foxhole” beliefs of a sort.

In positive-illusions situations such as those outlined above, the benefits accrue from misbelief directly – not merely from the systems that produce it. To return to the terminology we introduced earlier, such doxastic departures from reality – such apparent limitations of veridicality – are not culpable but entirely forgivable: design features, even. These beliefs are “Normal” in the capitalised, Millikanian sense. In such situations, we claim, we have our best candidates for evolved misbelief.

14. Ungrounded beliefs

Although “natural selection does not care about truth; it cares only about reproductive success” (Stich Reference Stich1990, p. 62), true beliefs can have instrumental value for natural selection – insofar as they facilitate reproductive success. In many cases (perhaps most), beliefs will be adaptive by virtue of their veridicality. The adaptiveness of such beliefs is not independent of their truth or falsity. On the other hand, the adaptiveness (or otherwise) of some beliefs is quite independent of their truth or falsity. Consider, again, supernatural belief: If belief in an omniscient, omnipotent deity is adaptive because it inhibits detectable selfish behaviour (as per the Johnson & Bering theory that we discussed in section 11), this will be the case whether or not such a being actually exists. If such a being does not exist, then we have adaptive misbelief. However, were such a being to suddenly pop into existence, the beliefs of a heretofore false believer would not become maladaptive – they would remain adaptive.

The misbeliefs that we have identified as sound candidates for adaptive misbelief are like the supernatural (mis)beliefs in the example above – although we claim that they were adaptive in themselves (not merely by-products of adaptively biased misbelief-producing systems), we do not claim that they were adaptive by virtue of their falsity: “Falseness itself could not be the point” (Millikan Reference Millikan2004, p. 86). It may be adaptive to believe that one's partner and one's children are more attractive (…etc.) than the average, but such adaptive beliefs are only adaptive misbeliefs, on our definition, if they happen to be false. Good grounds may arise for believing these things (success in beauty pageants, excessive attention from rivals, etc.), but such grounds will not render these beliefs any less adaptive. Their adaptiveness is independent of their truth or falsity. Any given adaptive misbeliever is thus an adaptive misbeliever because of contingent facts about the world – because her children are not actually as intelligent as she believes they are; because his prospects for recovery are not as good as he believes they are; and so forth. The upshot is that we do not expect adaptive misbeliefs to be generated by mechanisms specialised for the production of beliefs that are false per se. Instead, there will be evolved tendencies for forming specific ungrounded beliefs in certain domains. Where these beliefs are (contingently) false, we will see adaptive misbelief.

Dweck and colleagues (Dweck Reference Dweck1999; Blackwell et al. Reference Blackwell, Trzesniewski and Dweck2007) have shown a subtle instance of ungrounded belief (not necessarily false) that propels seemingly adaptive action. These authors distinguish two different “self-theories” of intelligence as part of the implicit “core beliefs” of adolescents: an “entity” theory (intelligence is a thing that you have a little or a lot of) and an “incremental” theory (intelligence is a malleable property that can develop). Those who hold an incremental theory are better motivated, work harder, and get better grades; and if students are taught an incremental theory in an intervention, they show significant improvement – and significantly more than that of a control group that is also given extra help but without the incremental theory. In fact, if students are told (truly or falsely) that they are particularly intelligent (intelligence is an entity and they have quite a lot of it), they actually do worse than if not told this. Note that these results are independent of the issue of whether or not an entity theory or an incremental theory is closer to the truth (or the truth about particular students). So regardless of whether one's intelligence is malleable, a belief that one's intelligence is malleable seems to have a strong positive effect on one's motivation and performance. It is tempting to conjecture that evolution has discovered this general tendency and exploited it: Whenever a belief about a desirable trait is “subjective” (Myers Reference Myers2002) and not likely to be rudely contradicted by experience, evolution should favour a disposition to err on the benign side, whatever it is, as this will pay dividends at little or no cost. Such an evolved bias could have the effect of instilling a host of unrealistically positive beliefs about oneself or about the vicissitudes to be encountered in the environment. What would hold this tendency in check, preventing people from living in fantasy worlds of prowess and paradise? As usual, the tendency should be self-limiting, with rash overconfidence leading to extinction in the not very long run (see Baumeister [Reference Baumeister1989] regarding the “optimal margin of illusion”).

If psychologists like Dweck can discover and manipulate these core beliefs today, our ancestors, with little or no theory or foresight, could have stumbled onto manipulations of the same factors and been amply rewarded by the effects achieved, turning their children into braver, more confident warriors, more trustworthy allies, more effective agents in many dimensions. Cultural evolution can have played the same shaping and pruning role as genetic evolution, yielding adaptations that pay for themselves – as all adaptations must – in the differential replication of those who adopt the cultural items, or in the differential replication of the cultural items themselves (Dawkins Reference Dawkins2006b; Dennett Reference Dennett1995a), or both. This in turn would open the door to gene-culture co-evolution, such as has been demonstrated with lactose tolerance in human lineages with a tradition of dairy herding (Beja-Pereira et al. Reference Beja-Pereira, Luikart, England, Bradley, Jann, Bertorelle, Chamberlain, Nunes, Metodiev, Ferrand and Erhardt2003; Feldman & Cavalli-Sforza Reference Feldman, Cavalli-Sforza and Feldman1989; Holden & Mace Reference Holden and Mace1997). Culturally evolved practices of inculcation could then create selective forces favouring those genetic variants that most readily responded to the inculcation, creating a genetically transmitted bias, a heightened susceptibility to those very practices (Dennett Reference Dennett2006a; McClenon Reference McClenon2002).