Breaking the Spell

by Daniel C. Dennett

  …we can say to them: “If standards of justice and truth are internal to each culture, you can have no objection to our characterization of you as war criminals. For just as our standards can have no application to you, your standards can have no application to us. We are as correct in proclaiming your evil in our culture as you are correct in proclaiming your uprightness in yours. But your very assertion that we have misunderstood you undermines this claim. It presupposes common values of truth and justice that we are somehow obligated to recognize. And on that ground we are prepared to argue for your wickedness.” [p. 148]

  This plea may fall on deaf ears, but if so, then there really are objective grounds for a verdict of irrationality: they are making a mistake that they themselves have no grounds to defend to themselves, and that we need not respect in deference.

  Cultural evolution has given us the thinking tools to create our societies and all their edifices and perspectives, and Balkin sees that these thinking tools—which he calls cultural software—are inevitably both liberating and constraining, both empowering and limiting. When our brains come to be inhabited by memes that have evolved under earlier selection pressures, our ways of thinking are restricted just as surely as our ways of talking and hearing are restricted when we learn our mother tongue. But the reflexivity that has evolved in human culture, the trick of thinking about thinking and representing our representations, makes all the restrictions temporary and revisable. As soon as we recognize that, we are ready to adopt what Balkin calls the ambivalent conception of ideology which avoids Mannheim’s paradox: “A subject constituted by cultural software is thinking about the cultural software that constitutes her. It is important to recognize that this recursion in and of itself involves no contradiction, anomaly, or logical difficulty” (pp. 127–28). Balkin insists, “Ideological critique does not stand above other forms of knowledge creation or acquisition. It is not a master form of knowing” (p. 134). This book is intended to be an instance of just such an ecumenical effort, relying on the respect for truth and the tools of truth-finding to provide a shared pool of knowledge from which we can work together toward mutually comprehended and accepted visions of what is good and what is just. The idea is not to bulldoze people with science, but to get them to see that things they already know, or could know, have implications for how they should want to respond to the issues under discussion.

  APPENDIX C The Bellboy and the Lady Named Tuck

  For years, Dan Sperber and his colleagues Scott Atran and Pascal Boyer have expressed their skepticism about the utility of the meme’s-eye perspective. First, let me try to give their main objections a clear expression, before saying why they have not persuaded me, in spite of what I have learned from them. This is my own summary of their position:

  It is obvious that cultural items (ideas, designs, methods, behaviors…) have population explosions and extinctions, and that there are large noncoincidental family resemblances between such items and the models that inspire them or from which they are otherwise descended. But the phenomenon of transmission in most but not quite all these cases is not the sort of high-fidelity copying that the gene model requires. The cause of a new instance is not copying at all: “The cause may merely trigger the production of a similar effect” (Sperber, 2000, p. 169). Thus produced, the similarities between instances are not like the similarities between genes and hence require a different sort of Darwinian explanation. Culture evolves, but not strictly by descent with modification. And it is true that there are some few memes meeting Dawkins’s specifications, such as chain letters, but such true memes play a relatively negligible role in the dynamics of cultural evolution (Sperber, 2000, p. 163). It is better to concentrate instead on the constraints and biases discernible in the psychological mechanisms that people share (Atran, 2002, pp. 237–38; Boyer, 2001, pp. 35–40).

  My main reply to this objection is to be found in appendix A, “The New Replicators.” Here I will expand on that reply by concentrating on the word italicized above: “instead.” I want to challenge the Sperberians’ conviction that they need to turn their backs on memes in order to study the constraints and biases of psychology. Atran, for instance, complains that the memetic approach is “mind-blind” (2002, pp. 241ff.) in that it ignores the detailed role of specific psychological mechanisms in shaping cultural items that proliferate. This is not at all an obvious point of disagreement, since Atran agrees that there is differential proliferation of cultural items. It is tempting to see the dispute as an artifact of miscommunication, with (some) memeticists promising too much and antimemeticists taking them at their word. As I note at the end of appendix A, memetics does not replace or pre-empt psychology any more than population genetics replaces or pre-empts ecology. (Is population genetics environmentblind? Yes, in general, and none the worse for it, since its models typically don’t go into the details of how and why there are selective pressures in the environment; they just show how the effects of those selective forces, whatever they are, will be manifested in populations over time as migrations and births and deaths take their toll. To get a whole biological explanation, you still need the ecology, and for a whole cultural explanation, memeticists still need psychology—though they may deny it in the throes of partisanship.)

  Boyer expresses the Sperberian objection in similar terms, but in spite of his stated opposition to memes, he often cannot resist couching his points in terms of differential replication. Indeed, his theory has been summarized by one sympathetic commentator as the thesis that “religion can primarily be understood as the systematic exploitation of mundane psychological systems by especially virulent strains of cultural concepts” (Bering, 2004, p. 126). “Virulent” is not quite the word Bering is seeking, since its (dictionary) connotations are all negative; “prolific” or “fit” would be a more accurate summary of Boyer’s thesis, since Boyer is careful to be neutral on the issue of whether religion is a good or bad accompaniment to human life, but, leaving that aside, it seems that Bering would include Boyer among the memeticists in spite of his disclaimers. So why couldn’t we just encourage Boyer and Atran and Sperber to concentrate on the selective forces provided by psychology, which they do so well, leaving the (trivial?) unificatory work to the memeticists down the corridor?

  But there is more to be said. We want to conceive of cultural evolution in terms of memes and in terms of the constraints of psychology—and the further constraints that emerge from the earlier interaction of memes and those very constraints! Consider an experiment we might do inspired by the research on “urban legends” by Heath, Bell, and Sternberg (2001). Did you hear about the bellboy who was caught on surveillance video putting the hotel guests’ toothbrushes up his…? How about the driver who heard a thump and when he stopped his car, many miles later, found the body of a baby embedded in the grillework of his car? Noting that many of the most popular urban legends involve shockingly disgusting tales, these researchers investigated the role of disgust in heightening the likelihood of transmission of a wide variety of urban legends. They provided competing “alleles” (alternative tellings) of each story and found that, sure enough, the more disgusting versions traveled better. Alas, they didn’t measure actual transmission, just their subjects’ convictions about how likely they were to repeat the stories. Research is expensive. But thought experiments are cheap, so let us imagine an experiment that would nicely illustrate the Sperberians’ point—and why I don’t find it a good argument against the memes approach.

  Suppose we concoct a thousand different urban legends—new ones, not yet circulating on the World Wide Web—and carefully plant them in ten thousand different hearers, one to a customer, each story going to ten hearers. We try to give these meme candidates “radioactive tags” by including telltale details in each planted version, along the lines of “Did you hear about the Brazilian taxi
-driver who…” And suppose we also spend lots of money tracking these trajectories, by hiring armies of private detectives to eavesdrop on our initial subjects, tapping their phones, and so forth (another virtue of thought experiments—you don’t have to clear them with your university’s internal review board or the police!), so that we get quite a lot of good data about which stories evaporate after a single telling, which actually get transmitted, in what words. The Sperberians’ dream result would be that we came up with…zilch! Almost all our radioactive tags would disappear, and all that would remain of the thousand different stories would be seven (say) stories that kept getting reinvented, time and again, because these seven stories were the only ones that tickled all the innate psychological constraints. When we looked at the lineages, we would see that, say, a hundred initially very different stories had all converged eventually on a single tale, the closest “attractor” in urban-legend space. Sometimes a story would be gradually modified in the direction of the favored attractor, but if the hearer already knew that tale, a new story might end abruptly in a cul-de-sac: “Hey, interesting. That reminds me—have you heard about the guy who…?”

  If this were the result, we would see that all the content in the urban legends that prevailed over time was already implicit in the psychology of the hearers and tellers, and virtually none was replicated faithfully from the initial stories. Here is Atran’s way of expressing the point:

  In genetic evolution there is only “weak selection” in the sense that there are no strong determinants of directional change. As a result, the cumulative effects of small mutations (on the order of one in a million) can lead to stable directional change. By contrast, in cultural evolution, there is very “strong selection” in the sense that modularized expectations can powerfully constrain transmitted information into certain channels and not others. As a result, despite frequent “error,” “noise,” and “mutation” in socially transmitted information, the messages tend to be steered (snapped back or launched forward) into cognitively stable paths. Cognitive modules, not memes themselves, enable the cultural canalization of beliefs and practices. [2002, p. 248]

  It would be almost as if we each have a CD in our brains with a few (dozen? hundred?) urban legends recorded on it; whenever we hear a close approximation to one of these urban legends, this triggers the CD to go to that track and play it—“triggered production,” not imitation of what we’ve heard. (This is suggested by Sperber’s “theoretical example” of the sound recorders [2000, p. 169].) That extreme null result is unlikely, of course, and if some content did get replicated from host to host, those who were infected by it would set up a new constraint on the fate of whatever urban legends they heard next. Cultural canalization can be due as much to prior cultural exposure as to one’s underlying cognitive modules. Perhaps, if you haven’t heard the one about the Chinese midget, you replicate the one about the boy with the pet gerbil and pass it along more or less intact, and if you have, you tend to merge them into something that eventually emerges as the one about the policewoman and the gerbil, and so forth. To investigate the interaction between contents culturally transmitted and constraints that are shared independently of culture, you really have to track the replication of memes—as best you can. Nobody said it was a practical research program in most instances.

  A remarkable instance of this occurred in the preparation of this book. One of the readers of the penultimate draft noticed a typographical error in chapter 2, and since it was repeated in the bibliography, it occurred to him that I might miss it: Gould’s 1999 book is Rocks of Ages, he told me, but I had written it Rock of Ages. My first reaction was frank disbelief. I thought my reader was making the mistake; the first word of Gould’s book couldn’t be “Rocks,” could it? I had read the book, and noted his plays on words (the paleontologist studies the ages of rocks, while…) but had completely missed his putting the mutation in his title, because the hymn title was so well branded into my memory! I had to check the book for myself, and, sure enough, the title is Rocks of Ages, but then I hopped on the Web to see if I was alone in making the error. On March 23, 2005, there were approximately as many Google citations for “Gould ‘Rock of Ages’” (3, 860) as for “Gould ‘Rocks of Ages’” (3, 950), and although many of the former entries proved to have both the correct title of Gould’s book and the hymn title, among the entries with the title misspelled were reviews of the book, and discussions of the book, both positive and negative. To casual inspection, there didn’t seem to be any obvious pattern to the errors, but here is a fine elementary project in computational memetics for anybody who wants to dig deeper. There is sure to be an interesting story to be told about how often this error has crept in by mutation and who has copied whose error. (See Dawkins’s discussion [1989, pp. 325–29] of a similar transcription error of a title, and an introduction to the methods of memetics using the resources of the Scientific Citation Index.)

  In addition to having the genetically evolved mechanisms or modules beloved by evolutionary psychologists, our brains are packed with culturally transmitted mechanisms of every imaginable sort, and the presence or absence of these sets up immunities and receptivities in hosts just as powerful as—or even more powerful than—the constraints exhibited by the underlying machinery. In his chapter against memes, Atran quotes me on this topic, but misses the point I was trying to make. I had said that the structure of Chinese and Korean minds is “dramatically different” from that of American or French minds (Dennett, 1995b, p. 365), and Atran supposes (2002, p. 258) that I am trying to make a subtle point about how people with different native tongues will interpret draw ings or attribute causation or blame in different scenarios. He cites experiments in which people from different cultural groups respond quite similarly in a variety of circumstances designed by psychologists to elicit such differences. But I had something much simpler and more obvious in mind: People with Chinese minds won’t laugh at, or remember, or repeat, jokes told in English! (A few years ago, the brilliant songwriter and singer Lyle Lovett released an album entitled Joshua Judges Ruth. I found that in general my friends didn’t get it; I’d ask them what Lovett’s next album might be entitled and none of them replied, “First and Second Samuel?”—which was the first thing that would pop into my head, thanks to Sunday-school drill more than half a century before.) And just as we can be quite sure that jokes told in French have a hard time getting spread in Anglophone neighborhoods, we can be quite sure that a person’s political views, and knowledge of art (or quantum physics, or sexual practices), would provide strong constraints or biases on his or her receptivity and eagerness to transmit various candidate memes. For instance, to my way of thinking, one of the funniest limericks I have ever heard is the following, which you will find funny only if you’ve heard lots of limericks:

  There was a young lady named Tuck

  Who had the most terrible luck:

  She went out in a punt,

  And fell over the front,

  And was bit on the leg by a duck.

  I couldn’t resist transmitting it to you. Who will transmit it further? That depends a lot on what other memes infect your brain, and the brains of those you talk to. In the complex world of cultural transmission, the patterns that are directly due to fixed features of human psychology will perhaps not loom particularly large. So it seems to me that those who follow Sperber in his opposition to memes are making points that can better be made in the language of memes: one of the things they are saying, for instance, is that convergent evolution plays such a dominating role in cultural evolution that the transmission of design by actual descent through cultural lineages is much less of a factor in accounting for observed similarities than the shaping of design by selective forces. This is often very plausible, and can be investigated in any case. But we should als
o be alert to the possibility that many of the similarities between, say, Islam and Christianity may be due to their common Abrahamic ancestor religion rather than to their each having adjusted to similar found conditions in their human adherents.

  APPENDIX D Kim Philby as a Real Case of Indeterminacy of Radical Interpretation

  Philosophers have spent decades dreaming up thought experiments designed to prove or disprove W.V.O. Quine’s (1960) principle of the indeterminacy of radical translation: the surprising claim that in principle there could be two different ways of translating one natural language into another natural language and no evidence at all about which one was the right way to translate the language. (Quine insisted that in that case there wouldn’t be a right way; each way would be as good as the other, and there would be no further fact of the matter.) It seems deeply unlikely at first that this would be possible. Couldn’t a well-informed bilingual, for instance, always tell which of two competing translations of a sentence in one of his languages was the better translation into the other? How could there not be plenty of evidence in favor of one of the two translations?

  If you think the resolution is obvious, you haven’t read, or understood, the voluminous philosophical literature on this curious puzzle. A good place to start, after reading Quine’s masterpiece, Word and Object (1960), would be the special 1974 issue of Synthese devoted to a University of Connecticut conference on intentionality, language, and translation, in which Quine took on his most distinguished opponents and left them, and the issue, unresolved, which is where the issue stands today (Quine, 1974a, b).