Viruses of the Mind
Richard Dawkins

1 Duplication Fodder

A human child is shaped by evolution to soak up the culture of her people. Most obviously, she learns the essentials of their language in a matter of months. A large dictionary of words to speak, an encyclopedia of information to speak about, complicated syntactic and semantic rules to order the speaking, are all transferred from older brains into hers well before she reaches half her adult size. When you are pre-programmed to absorb useful information at a high rate, it is hard to shut out pernicious or damaging information at the same time. With so many mindbytes to be downloaded, so many mental codons to be replicated, it is no wonder that child brains are gullible, open to almost any suggestion, vulnerable to subversion, easy prey to Moonies, Scientologists and nuns. Like immune-deficient patients, children are wide open to mental infections that adults might brush off without effort.

DNA, too, includes parasitic code. Cellular machinery is extremely good at copying DNA. Where DNA is concerned, it seems to have an eagerness to copy, seems eager to be copied. The cell nucleus is a paradise for DNA, humming with sophisticated, fast, and accurate duplicating machinery.

Cellular machinery is so friendly towards DNA duplication that it is small wonder cells play host to DNA parasites --- viruses, viroids, plasmids and a riff-raff of other genetic fellow travelers. Parasitic DNA even gets itself spliced seamlessly into the chromosomes themselves. ``Jumping genes'' and stretches of ``selfish DNA'' cut or copy themselves out of chromosomes and paste themselves in elsewhere. Deadly oncogenes are almost impossible to distinguish from the legitimate genes between which they are spliced. In evolutionary time, there is probably a continual traffic from ``straight'' genes to ``outlaw,'' and back again (Dawkins, 1982). DNA is just DNA. The only thing that distinguishes viral DNA from host DNA is its expected method of passing into future generations. ``Legitimate'' host DNA is just DNA that spires to pass into the next generation via the orthodox route of sperm or egg. ``Outlaw'' or parasitic DNA is just DNA that looks to a quicker, less cooperative route to the future, via a squeezed droplet or a smear of blood, rather than via a sperm or egg.

For data on a floppy disc, a computer is a humming paradise just as cell nuclei hum with eagerness to duplicate DNA. Computers and their associated disc and tape readers are designed with high fidelity in mind. As with DNA molecules, magnetized bytes don't literally ``want'' to be faithfully copied. Nevertheless, you can write a computer program that takes steps to duplicate itself. Not just duplicate itself within one computer but spread itself to other computers. Computers are so good at copying bytes, and so good at faithfully obeying the instructions contained in those bytes, that they are sitting ducks to self-replicating programs: wide open to subversion by software parasites. Any cynic familiar with the theory of selfish genes and memes would have known that modern personal computers, with their promiscuous traffic of floppy discs and e-mail links, were just asking for trouble. The only surprising thing about the current epidemic of computer viruses is that it has been so long in coming.

2 Computer Viruses: a Model for an Informational Epidemiology

``Memes now spread around the world at the speed of light, and replicate at rates that make even fruit flies and yeast cells look glacial in comparison. They leap promiscuously from vehicle to vehicle, and from medium to medium, and are proving to be virtually unquarantinable'' (Dennett 1990, p.131). Viruses aren't limited to electronic media such as disks and data lines. On its way from one computer to another, a virus may pass through printing ink, light rays in a human lens, optic nerve impulses and finger muscle contractions. A computer fanciers' magazine that printed the text of a virus program for the interest of its readers has been widely condemned. Indeed, such is the appeal of the virus idea to a certain kind of puerile mentality (the masculine gender is used advisedly), that publication of any kind of ``how to'' information on designing virus programs is rightly seen as an irresponsible act.

I am not going to publish any virus code. But there are certain tricks of effective virus design that are sufficiently well known, even obvious, that it will do no harm to mention them, as I need to do to develop my theme. They all stem from the virus's need to evade detection while it is spreading.

A virus that clones itself too prolifically within one computer will soon be detected because the symptoms of clogging will become too obvious to ignore. For this reason many virus programs check, before infecting a system, to make sure that they are not already on that system. Incidentally, this opens the way for a defense against viruses that is analogous to immunization. In the days before a specific anti-virus program was available, I myself responded to an early infection of my own hard disk by means of a crude ``vaccination.'' Instead of deleting the virus that I had detected, I simply disabled its coded instructions, leaving the ``shell'' of the virus with its characteristic external ``signature'' intact. In theory, subsequent members of the same virus species that arrived in my system should have recognized the signature of their own kind and refrained from trying to double-infect. I don't know whether this immunization really worked, but in those days it probably was worth while ``gutting'' a virus and leaving a shell like this, rather than simply removing it lock, stock and barrel. Nowadays it is better to hand the problem over to one of the professionally written anti-virus programs.

A virus that is too virulent will be rapidly detected and scotched. A virus that instantly and catastrophically sabotages every computer in which it finds itself will not find itself in many computers. It may have a most amusing effect on one computer ---- erase an entire doctoral thesis or something equally side-splitting --- but it won't spread as an epidemic.

Some viruses, therefore, are designed to have an effect that is small enough to be difficult to detect, but which may nevertheless be extremely damaging. There is one type, which, instead of erasing disk sectors wholesale, attacks only spreadsheets, making a few random changes in the (usually financial) quantities entered in the rows and columns. Other viruses evade detection by being triggered probabilistically, for example erasing only one in 16 of the hard disks infected. Yet other viruses employ the time-bomb principle. Most modern computers are ``aware'' of the date, and viruses have been triggered to manifest themselves all around the world, on a particular date such as Friday 13th or April Fool's Day. From the parasitic point of view, it doesn't matter how catastrophic the eventual attack is, provided the virus has had plenty of opportunity to spread first (a disturbing analogy to the Medawar/Williams theory of ageing: we are the victims of lethal and sub-lethal genes that mature only after we have had plenty of time to reproduce (Williams, 1957)). In defense, some large companies go so far as to set aside one ``miner's canary'' among their fleet of computers, and advance its internal calendar a week so that any time-bomb viruses will reveal themselves prematurely before the big day.

Again predictably, the epidemic of computer viruses has triggered an arms race. Anti-viral software is doing a roaring trade. These antidote programs -- ``Interferon,'' ``Vaccine,'' ``Gatekeeper'' and others --- employ a diverse armory of tricks. Some are written with specific, known and named viruses in mind. Others intercept any attempt to meddle with sensitive system areas of memory and warn the user.

The virus principle could, in theory, be used for non-malicious, even beneficial purposes. Thimbleby (1991) coins the phrase ``liveware'' for his already-implemented use of the infection principle for keeping multiple copies of databases up to date. Every time a disk containing the database is plugged into a computer, it looks to see whether there is already another copy present on the local hard disk. If there is, each copy is updated in the light of the other. So, with a bit of luck, it doesn't matter which member of a circle of colleagues enters, say, a new bibliographical citation on his personal disk. His newly entered information will readily infect the disks of his colleagues (because the colleagues promiscuously insert their disks into one another's computers) and will spread like an epidemic around the circle. Thimbleby's liveware is not entirely virus-like: it could not spread to just anybody's computer and do damage. It spreads data only to already-existing copies of its own database; and you will not be infected by liveware unless you positively opt for infection.

Incidentally, Thimbleby, who is much concerned with the virus menace, points out that you can gain some protection by using computer systems that other people don't use. The usual justification for purchasing today's numerically dominant computer is simply and solely that it is numerically dominant. Almost every knowledgeable person agrees that, in terms of quality and especially user-friendliness, the rival, minority system is superior. Nevertheless, ubiquity is held to be good in itself, sufficient to outweigh sheer quality. Buy the same (albeit inferior) computer as your colleagues, the argument goes, and you'll be able to benefit from shared software, and from a generally large circulation of available software. The irony is that, with the advent of the virus plague, ``benefit'' is not all that you are likely to get. Not only should we all be very hesitant before we accept a disk from a colleague. We should also be aware that, if we join a large community of users of a particular make of computer, we are also joining a large community of viruses --- even, it turns out, disproportionately larger.

Returning to possible uses of viruses for positive purposes, there are proposals to exploit the ``poacher turned gamekeeper'' principle, and ``set a thief to catch a thief.'' A simple way would be to take any of the existing anti-viral programs and load it, as a ``warhead,'' into a harmless self-replicating virus. From a ``public health'' point of view, a spreading epidemic of anti-viral software could be especially beneficial because the computers most vulnerable to malicious viruses --- those whose owners are promiscuous in the exchange of pirated programs --- will also be most vulnerable to infection by the healing anti-virus. A more penetrating anti-virus might --- as in the immune system --- ``learn'' or ``evolve'' an improved capacity to attack whatever viruses it encountered.

I can imagine other uses of the computer virus principle which, if not exactly altruistic, are at least constructive enough to escape the charge of pure vandalism. A computer company might wish to do market research on the habits of its customers, with a view to improving the design of future products. Do users like to choose files by pictorial icon, or do they opt to display them by textual name only? How deeply do people nest folders (directories) within one another? Do people settle down for a long session with only one program, say a word processors, or are they constantly switching back and forth, say between writing and drawing programs? Do people succeed in moving the mouse pointer straight to the target, or do they meander around in time-wasting hunting movements that could be rectified by a change in design?

The company could send out a questionnaire asking all these questions, but the customers that replied would be a biased sample and, in any case, their own assessment of their computer-using behavior might be inaccurate. A better solution would be a market-research computer program. Customers would be asked to load this program into their system where it would unobtrusively sit, quietly monitoring and tallying key-presses and mouse movements. At the end of a year, the customer would be asked to send in the disk file containing all the tallyings of the market-research program. But again, most people would not bother to cooperate and some might see it as an invasion of privacy and of their disk space.

The perfect solution, from the company's point of view, would be a virus. Like any other virus, it would be self-replicating and secretive. But it would not be destructive or facetious like an ordinary virus. Along with its self-replicating booster it would contain a market-research warhead. The virus would be released surreptitiously into the community of computer users. Just like an ordinary virus it would spread around, as people passed floppy disks and e-mail around the community. As the virus spread from computer to computer, it would build up statistics on users behavior, monitored secretly from deep within a succession of systems. Every now and again, a copy of the viruses would happen to find its way, by normal epidemic traffic, back into one of the company's own computers. There it would be debriefed and its data collated with data from other copies of the virus that had come ``home.''

Looking into the future, it is not fanciful to imagine a time when viruses, both bad and good, have become so ubiquitous that we could speak of an ecological community of viruses and legitimate programs coexisting in the silicosphere. At present, software is advertised as, say, ``Compatible with System 7.'' In the future, products may be advertised as ``Compatible with all viruses registered in the 1998 World Virus Census; immune to all listed virulent viruses; takes full advantage of the facilities offered by the following benign viruses if present...'' Word-processing software, say, may hand over particular functions, such as word-counting and string-searches, to friendly viruses burrowing autonomously through the text.

Looking even further into the future, whole integrated software systems might grow, not by design, but by something like the growth of an ecological community such as a tropical rain-forest. Gangs of mutually compatible viruses might grow up, in the same way as genomes can be regarded as gangs of mutually compatible genes (Dawkins, 1982). Indeed, I have even suggested that our genomes should be regarded as gigantic colonies of viruses (Dawkins, 1976). Genes cooperate with one another in genomes because natural selection has favored those genes that prosper in the presence of the other genes that happen to be common in the gene pool. Different gene pools may evolve towards different combinations of mutually compatible genes. I envisage a time when, in the same kind of way, computer viruses may evolve towards compatibility with other viruses, to form communities or gangs. But then again, perhaps not! At any rate, I find the speculation more alarming than exciting.

At present, computer viruses don't strictly evolve. They are invented by human programmers, and if they evolve they do so in the same weak sense as cars or aeroplanes evolve. Designers derive this year's car as a slight modification of last year's car, and then may, more or less consciously, continue a trend of the last few years --- further flattening of the radiator grill or whatever it may be. Computer virus designers dream up ever more devious tricks for outwitting the programmers of anti-virus software. But computer viruses don't --- so far --- mutate and evolve by true natural selection. They may do so in the future. Whether they evolve by natural selection, or whether their evolution is steered by human designers, may not make much difference to their eventual performance. By either kind of evolution, we expect them to become better at concealment, and we expect them to become subtly compatible with other viruses that are at the same time prospering in the computer community.

DNA viruses and computer viruses spread for the same reason: an environment exists in which there is machinery well set up to duplicate and spread them around and to obey the instructions that the viruses embody. These two environments are, respectively, the environment of cellular physiology and the environment provided by a large community of computers and data-handling machinery. Are there any other environments like these, any other humming paradises of replication?

3 The Infected Mind

Think about the two qualities that a virus, or any sort of parasitic replicator, demands of a friendly medium,. the two qualities that make cellular machinery so friendly towards parasitic DNA, and that make computers so friendly towards computer viruses. These qualities are, firstly, a readiness to replicate information accurately, perhaps with some mistakes that are subsequently reproduced accurately; and, secondly, a readiness to obey instructions encoded in the information so replicated.

Cellular machinery and electronic computers excel in both these virus-friendly qualities. How do human brains match up? As faithful duplicators, they are certainly less perfect than either cells or electronic computers. Nevertheless, they are still pretty good, perhaps about as faithful as an RNA virus, though not as good as DNA with all its elaborate proofreading measures against textual degradation. Evidence of the fidelity of brains, especially child brains, as data duplicators is provided by language itself. Shaw's Professor Higgins was able by ear alone to place Londoners in the street where they grew up. Fiction is not evidence for anything, but everyone knows that Higgins's fictional skill is only an exaggeration of something we can all down. Any American can tell Deep South from Mid West, New England from Hillbilly. Any New Yorker can tell Bronx from Brooklyn. Equivalent claims could be substantiated for any country. What this phenomenon means is that human brains are capable of pretty accurate copying (otherwise the accents of, say, Newcastle would not be stable enough to be recognized) but with some mistakes (otherwise pronunciation would not evolve, and all speakers of a language would inherit identically the same accents from their remote ancestors). Language evolves, because it has both the great stability and the slight changeability that are prerequisites for any evolving system.

The second requirement of a virus-friendly environment --- that it should obey a program of coded instructions --- is again only quantitatively less true for brains than for cells or computers. We sometimes obey orders from one another, but also we sometimes don't. Nevertheless, it is a telling fact that, the world over, the vast majority of children follow the religion of their parents rather than any of the other available religions. Instructions to genuflect, to bow towards Mecca, to nod one's head rhythmically towards the wall, to shake like a maniac, to ``speak in tongues'' --- the list of such arbitrary and pointless motor patterns offered by religion alone is extensive --- are obeyed, if not slavishly, at least with some reasonably high statistical probability.

Less portentously, and again especially prominent in children, the ``craze'' is a striking example of behavior that owes more to epidemiology than to rational choice. Yo-yos, hula hoops and pogo sticks, with their associated behavioral fixed actions, sweep through schools, and more sporadically leap from school to school, in patterns that differ from a measles epidemic in no serious particular. Ten years ago, you could have traveled thousands of miles through the United States and never seen a baseball cap turned back to front. Today, the reverse baseball cap is ubiquitous. I do not know what the pattern of geographical spread of the reverse baseball cap precisely was, but epidemiology is certainly among the professions primarily qualified to study it. We don't have to get into arguments about ``determinism''; we don't have to claim that children are compelled to imitate their fellows' hat fashions. It is enough that their hat-wearing behavior, as a matter of fact, is statistically affected by the hat-wearing behavior of their fellows.

Trivial though they are, crazes provide us with yet more circumstantial evidence that human minds, especially perhaps juvenile ones, have the qualities that we have singled out as desirable for an informational parasite. At the very least the mind is a plausible candidate for infection by something like a computer virus, even if it is not quite such a parasite's dream-environment as a cell nucleus or an electronic computer.

It is intriguing to wonder what it might feel like, from the inside, if one's mind were the victim of a ``virus.'' This might be a deliberately designed parasite, like a present-day computer virus. Or it might be an inadvertently mutated and unconsciously evolved parasite. Either way, especially if the evolved parasite was the memic descendant of a long line of successful ancestors, we are entitled to expect the typical ``mind virus'' to be pretty good at its job of getting itself successfully replicated.

Progressive evolution of more effective mind-parasites will have two aspects. New ``mutants'' (either random or designed by humans) that are better at spreading will become more numerous. And there will be a ganging up of ideas that flourish in one another's presence, ideas that mutually support one another just as genes do and as I have speculated computer viruses may one day do. We expect that replicators will go around together from brain to brain in mutually compatible gangs. These gangs will come to constitute a package, which may be sufficiently stable to deserve a collective name such as Roman Catholicism or Voodoo. It doesn't too much matter whether we analogize the whole package to a single virus, to each one of the component parts to a single virus. The analogy is not that precise anyway, just as the distinction between a computer virus and a computer worm is nothing to get worked up about. What matters is that minds are friendly environments to parasitic, self-replicating ideas or information, and that minds are typically massively infected.

Like computer viruses, successful mind viruses will tend to be hard for their victims to detect. If you are the victim of one, the chances are that you won't know it, and may even vigorously deny it. Accepting that a virus might be difficult to detect in your own mind, what tell-tale signs might you look out for? I shall answer by imaging how a medical textbook might describe the typical symptoms of a sufferer (arbitrarily assumed to be male).

1. The patient typically finds himself impelled by some deep, inner conviction that something is true, or right, or virtuous: a conviction that doesn't seem to owe anything to evidence or reason, but which, nevertheless, he feels as totally compelling and convincing. We doctors refer to such a belief as ``faith.''

2. Patients typically make a positive virtue of faith's being strong and unshakable, in spite of not being based upon evidence. Indeed, they may fell that the less evidence there is, the more virtuous the belief (see below).

This paradoxical idea that lack of evidence is a positive virtue where faith is concerned has something of the quality of a program that is self-sustaining, because it is self-referential (see the chapter ``On Viral Sentences and Self-Replicating Structures'' in Hofstadter, 1985). Once the proposition is believed, it automatically undermines opposition to itself. The ``lack of evidence is a virtue'' idea could be an admirable sidekick, ganging up with faith itself in a clique of mutually supportive viral programs.

3. A related symptom, which a faith-sufferer may also present, is the conviction that ``mystery,'' per se, is a good thing. It is not a virtue to solve mysteries. Rather we should enjoy them, even revel in their insolubility.

Any impulse to solve mysteries could be serious inimical to the spread of a mind virus. It would not, therefore, be surprising if the idea that ``mysteries are better not solved'' was a favored member of a mutually supporting gang of viruses. Take the ``Mystery of Transubstantiation.'' It is easy and non-mysterious to believe that in some symbolic or metaphorical sense the eucharistic wine turns into the blood of Christ. The Roman Catholic doctrine of transubstantiation, however, claims far more. The ``whole substance'' of the wine is converted into the blood of Christ; the appearance of wine that remains is ``merely accidental,'' ``inhering in no substance'' (Kenny, 1986, p. 72). Transubstantiation is colloquially taught as meaning that the wine ``literally'' turns into the blood of Christ. Whether in its obfuscatory Aristotelian or its franker colloquial form, the claim of transubstantiation can be made only if we do serious violence to the normal meanings of words like ``substance'' and ``literally.'' Redefining words is not a sin, but, if we use words like ``whole substance'' and ``literally'' for this case, what word are we going to use when we really and truly want to say that something did actually happen? As Anthony Kenny observed of his own puzzlement as a young seminarian, ``For all I could tell, my typewriter might be Benjamin Disraeli transubstantiated....''

Roman Catholics, whose belief in infallible authority compels them to accept that wine becomes physically transformed into blood despite all appearances, refer to the ``mystery'' of transubstantiation. Calling it a mystery makes everything OK, you see. At least, it works for a mind well prepared by background infection. Exactly the same trick is performed in the ``mystery'' of the Trinity. Mysteries are not meant to be solved, they are meant to strike awe. The ``mystery is a virtue'' idea comes to the aid of the Catholic, who would otherwise find intolerable the obligation to believe the obvious nonsense of the transubstantiation and the ``three-in-one.'' Again, the belief that ``mystery is a virtue'' has a self-referential ring. As Hofstadter might put it, the very mysteriousness of the belief moves the believer to perpetuate the mystery.

An extreme symptom of ``mystery is a virtue'' infection is Tertullian's ``Certum est quia impossibile est'' (It is certain because it is impossible''). That way madness lies. One is tempted to quote Lewis Carroll's White Queen, who, in response to Alice's ``One can't believe impossible things'' retorted ``I daresay you haven't had much practice... When I was your age, I always did it for half-an-hour a day. Why, sometimes I've believed as many as six impossible things before breakfast.'' Or Douglas Adam's Electric Monk, a labor-saving device programmed to do your believing for you, which was capable of ``believing things they'd have difficulty believing in Salt Lake City'' and which, at the moment of being introduced to the reader, believed, contrary to all the evidence, that everything in the world was a uniform shade of pink. But White Queens and Electric Monks become less funny when you realize that these virtuoso believers are indistinguishable from revered theologians in real life. ``It is by all means to be believed, because it is absurd'' (Tertullian again). Sir Thomas Browne (1635) quotes Tertullian with approval, and goes further: ``Methinks there be not impossibilities enough in religion for an active faith.'' And ``I desire to exercise my faith in the difficultest point; for to credit ordinary and visible objects is not faith, but perswasion [sic].''

I have the feeling that something more interesting is going on here than just plain insanity or surrealist nonsense, something akin to the admiration we feel when we watch a ten-ball juggler on a tightrope. It is as though the faithful gain prestige through managing to believe even more impossible things than their rivals succeed in believing. Are these people testing --- exercising --- their believing muscles, training themselves to believe impossible things so that they can take in their stride the merely improbable things that they are ordinarily called upon to believe?

While I was writing this, the Guardian (July 29, 1991) fortuitously carried a beautiful example. It came in an interview with a rabbi undertaking the bizarre task of vetting the kosher-purity of food products right back to the ultimate origins of their minutest ingredients. He was currently