The Spandrels of Self-Deception: Prospects for a Biological Theory of a Mental Phenomenon

Abstract

Three puzzles about self-deception make this mental phenomenon an intriguing explanatory target. The first relates to how to define it without paradox; the second is about how to make sense of self-deception in light of the interpretive view of the mental that has become widespread in philosophy; and the third concerns why it exists at all. In this paper I address the first and third puzzles. First, I define self-deception. Second, I criticize Robert Trivers’ attempt to use adaptionist evolutionary psychology to solve the third puzzle (existence). Third, I sketch a theory to replace that of Trivers. Self-deception is not an adaptation, but a spandrel in the sense that Gould and Lewontin give the term: a byproduct of other features of human (cognitive) architecture.

Self-deception is so undeniable a fact of human life that if anyone tried to deny its existence, the proper response would be to accuse this person of it. (Allen Wood, 1988 )

Keywords:

• Adaptationism
• Evolutionary Psychology
• Stephen Jay Gould
• Richard Lewontin
• Self-Deception
• Spandrel
• Robert Trivers

Acknowledgements

For helpful exchanges on the subject matter of this paper, I’d like to thank: Alan Lloyd, Alfred Mele, John Perry, Erica Roeder, Elliott Sober, Kenneth Taylor, Robert Trivers, Avi Tuschman, the audience at the 2005 Columbia/NYU 5th Annual Graduate Conference in Philosophy, the audience at the Stanford Philosophy WIP Seminar, and two anonymous referees for Philosophical Psychology.

Notes

Notes

[1] Jean-Paul Sartre (1956, p. 89) raises a closely analogous paradox regarding his concept “bad faith.”

[2] I am not so concerned to deal with puzzles 1 and 2 in this paper, although I believe my next section on the concept of self-deception offers a start. The literature, however, is extensive. For work relevant to puzzle 1, see, e.g., McLaughlin (1988) and Rorty (1988). Davidson's own work (1982, 1985, 1998) is largely addressed at puzzle 2. See Johnston (1988) for a view that takes puzzle 2 as a reductio of the interpretive view.

[3] Mele's work may be viewed as an exception to this claim. However, he is more concerned with how self-deception comes about rather than with the question of why the capacity for it exists.

[4] Trivers also adopts the strategy of explaining self-deception as selected to aid deception in Trivers (1985) and in the forward to Dawkins (1976). I focus on Trivers (2000) because it is most recent.

[5] The first entry in the online Oxford English Dictionary for ‘misrepresentation’ is: “Wrong or incorrect representation of facts, statements, the character of a person, etc.”

[6] It may be further asked what I mean by ‘usual’ here, since if ‘usual’ just means non-self-deceptive then we are stuck with a circular definition. This problem, however, can be solved by saying that ‘usual’ just means ‘in absence of the kind of belief-influencing desire that constitutes the second point of commonality between the two cases’.

[7] Patten (2003) argues that a desire is not necessary for self-deception. On his view, an agent can become self-deceived about her own motives through mistakes caused by nonmotivated biases of the sort that cause one to make mistakes about someone else's motives. Such mistakes about oneself are certainly possible, but it is questionable whether in their nonmotivated form they should be viewed as self-deception. If there's no motivation involved in driving the epistemic failure, it might be more appropriate to call the resulting mistake an honest mistake.

[8] Of course, the content of the desire must be related to the content of the self-deceptively held belief, e.g., the cuckold in denial desires that his wife is faithful and this is what he self-deceptively believes, so in that case the relation of content is identity. But there are other possibilities besides identity. Nelkin (2002), for example, argues that the content of the motivating desire is to believe what ultimately comes to be self-deceptively believed. As my example of the quarterback indicates, I think this is one possibility. But how the contents of the two elements must be related is a subject of debate that wouldn’t be an appropriate focus for this article.

[9] There is disagreement about which of these two notions is the right analysis of self-deception. McLaughlin (1988) is explicit about the view that self-deception is on a continuum with wishful thinking, while Talbott (1995) adopts the “intentional” self-deception analysis. By defining the term in the inclusive way that I do, I avoid the controversy. Of course, I have independent reasons for adopting the analysis I do.

[10] For a discussion arguing that some sort of motivation is needed for a case of irrationality to be a case of self-deception, see Mele (2001, pp. 104–110), who argues against Martha Knight's (1988) purely cognitive explanation of certain forms of self-deception. In my view, it may be that other types of irrationality that are not motivated have etiologies that are similar to the etiology of what I call self-deception. That is, they will be spandrels of some of the same cognitive features of which the capacity for self-deception is a spandrel. This would make them “cousins” of self-deception without being precisely the same phenomenon as that in which most philosophers and psychologists who have discussed self-deception have been interested. Those who build motivation into their constitutive characterization of self-deception include: Audi (1982, 1988), Bach (1981), Davidson (1982, 1985, 1998), Funkhouser (2005), Gur and Sackeim (1979), Mele (2001), Nelkin (2002), Pears (1984), Rey (1988), and others. This paper is unfortunately not the place to argue conclusively that motivation must be included in the definition of self-deception.

[11] Mele (2001) calls this “twisted self-deception.”

[12] It is also not perfectly clear that Trivers is positing a self-deception module or modules, but his emphasis on modules on pages 116–117 make this the most likely interpretation.

[13] Depending on how one understands ‘deceive’, if one is truly self-deceived in having a certain belief, then propagating that belief to others will not count as deceiving, because one has the belief oneself. But I trust it will be clear what is meant in this context: ‘deceiving’ is standing in for ‘conveying beliefs that one would not under normal circumstances believe oneself ’. In this text, I often use ‘lie’ instead of ‘deceive’ for stylistic reasons.

[14] In presenting this argument to others, I’ve encountered the objection more than once that skittish animals like mice are counterexamples to the claim that having good information channels is generally adaptive. The objection is that a mouse is better adapted for believing any hovering shadow to be a hawk, even when it's not true that it is. The first thing to note is that, if this really is a counterexample, it's an exception to a rule that holds widely. The mouse's eyes, ears, nose, and whiskers send its brain information that on the whole is processed in a highly sophisticated and reliable way. In any case, the tendency towards overrepresentation of HAWK in the mouse's cognitive processing has an easy explanation. This is not so for self-deception, which can infect a very large variety of beliefs—ranging from beliefs about lovers to sports to politics—and often has destructive consequences. In any case, it is clear that Trivers has the same back ground assumption in mind when asking the following rhetorical question: “In trying to deal effectively with a complex, changing world, where is the benefit in misrepresenting reality to oneself ?” (p. 115). In other words, Trivers is in agreement with me in taking the subversion of internal information that goes on in self-deception as something to be explained. I’m arguing here that his explanation isn’t adequate.

[15] Ramachandran (1998, pp. 278–279) has an example in his endnote 3 on Trivers’ view that parallels my own. Ramachandran's example is of a chimp deceiving another chimp about the whereabouts of bananas, who, if he's self-deceived, won’t get the bananas either. Ramachandran says this seems to him to be an “internal contradiction” in Trivers’ theory, although he does go on to speculate how it might be resolved.

[16] It would be a weak objection to this point to say that it does not hold because there were not quarterbacks in the ancestral environment; whatever that environment was, it will not be that difficult to construct analogous cases (assuming the environment was social and people were at all disposed to get nervous for social reasons).

[17] See, e.g., Taylor & Brown (1988) and Taylor (1989, 1998). I’d like to thank one anonymous reviewer for encouraging me to address this issue.

[18] This is a general rule that has many exceptions—forms of behavior that seem aimed at obtaining discomfort, e.g., masochism. Nevertheless, the general rule holds widely and suffices for present explanatory purposes.

[19] These five features of human cognition are, of course, not at all new ideas. What I think is original, if anything, is the way I employ (3) and (4) to understand the phenomenon of self-deception ([2] and [5] have figured in other discussions) and the insight that the way such features work together to produce self-deception affects what sort of biological stance we may take on the phenomenon. My inspiration for (3) comes from Quine (1953) and Kuhn (1962). I realized that (4) may play a role in self-deception on reading Sober (1981, pp. 103ff.). (2) is discussed in Talbott (1995).

[20] The one feature I cite for which I was unable to find support in the form of experimental data is (4), the disposition to favor theories that are on the whole less complex. My thought on the matter is, however, motivated by the following kind of case. Suppose you’re a detective faced with one suspect who has a simple explanation for being near the scene of the crime and one with a complicated explanation. I think there's a natural inclination to give the simpler explanation more credit.

[21] This analysis leads to a possible connection between self-deception and deception in which the direction of explanation is the opposite of what Trivers gives (by this I mean that attempts to deceive can explain some cases of self-deception). Often in order to deceive, one must gather evidence in support of the deception. The biased accumulation of evidence can then lead the deceiver himself to be deceived.

[22] The “inclination” to favor less complex theories on the whole should not be read as satisfying the desire component of my definition of self-deception. The inclination is rather a cognitive tendency that can contribute to the violation of epistemic norms in self-deception. This means of course that feature (4) cannot by itself cause a phenomenon satisfying the definition of self-deception to come about, so other features, e.g., (1), (2) and/or (5), may be required.

[23] Of course, some of these features could combine to produce spandrel byproducts that do not count as self-deception. See note 10.

[24] I’d like to thank one anonymous reviewer for making me aware of the possibility of weaker forms of adaptationism about self-deception.